Values-Based Operational Decision-Making: A Practitioner's Framework for Measuring Virtue-Culture Alignment

Rob Davidson | Koch Center Own-Operate-Invest Fellowship | WashU Olin | May 2026

A Practitioner's Framework for Measuring Virtue-Culture Alignment

---

Abstract

Every business runs on two sets of values: the ones on the wall and the ones in the hallway. The gap between them is realized in coordination cost, disengagement, and rework; the firm runs less efficiently and leaves operating profit on the table. This paper introduces the Virtue-Culture Alignment Framework (VCAF), a methods-and-practice proposal that measures the gap between declared and practiced values, virtue by virtue. The employee narratives that produce the practiced-virtue score double as the forensic record of where the gap surfaces as operating contradictions, yielding a function-level coding that guides the intervention. The framework draws on Aristotle's treatment of virtue as habit. Its core assumption (Step 5): virtue-culture alignment is expected to reduce coordination friction, such that gains in the Ethical Alignment Index (EAI) correspond to gains in realized operating profit. The Ethical Output Value (EOV) is expressed primarily in Adjusted EBITDA, the measure most sensitive to cross-departmental coordination.

---

Contribution

VCAF is an operating diagnostic that measures declared-versus-practiced virtue gaps, routes those gaps to specific operating contradictions, and expresses the size of the issue in EBITDA-risk terms until evidence over time can validate predictive claims. The contribution is methodological and practical: a structured operator diagnostic, an intervention runbook tied to pre-committed metric anchors, and a financial translation pending validation. The framework's primary informational output is the function-targeted intervention plan; the dollar figure is the diligence translation that forces pre-commitment to what would count as the intervention working.

The instruments this framework builds on each stop short of the combination VCAF attempts. Kaptein's Corporate Ethical Virtues model validates multi-virtue diagnostics across dozens of studies but does not link scores to financial outcomes. Denison's Organizational Culture Survey and Gallup's Q12 connect culture to performance without resolving at the virtue level. Bloom and Van Reenen's World Management Survey delivers identification-credible evidence that management practice drives productivity, including randomized trials, but measures operational management rather than virtues and does not split declared from practiced. McKinsey's Organizational Health Index benchmarks organizational health at scale but requires infrastructure most mid-market firms do not have. The full instrument comparison is in Appendix G.

---

Why This Paper Exists

I grew up in a manufacturing family. My grandfather founded two steel companies in rural Southern Illinois, Red Bud Industries and The Material Works. Both still operate today.

When a downturn hit Red Bud Industries and margins got so thin the company couldn't make payroll, my grandfather ran every employee's check, put them in envelopes, and wrote letters telling people the company didn't have the money to distribute them. He asked that no one cash their check until the company recovered; if someone really needed it, they could come ask. They kept showing up. He'd been in the trenches next to them for years, pouring what he made back into the community; they'd watched him live what he asked of them. That's why the plant is still running today.

That story stuck with me. Culture is the differentiator. When it works, you get discretionary effort and people who solve problems before you know they exist. When it doesn't, you get compliance. And compliance will always be the ceiling.

I have spent the last decade building teams. When practiced culture matches what leaders declare, performance compounds. When the two diverge, performance stalls and the firm pays for the gap in margin no single line item captures. This paper builds a tool to measure that gap at virtue resolution, because organization-level culture scores tell a leader something is wrong without telling them which management system is producing it. I am building it for the leaders I work with, for the communities their firms are rooted in, and so my son grows up watching what a values-explicit business looks like before he reads about one.

---

Values Versus Virtues

Values are declarations: what a company says it believes. Virtues are values made real through repetition: behavioral norms that reduce coordination failure under pressure, the habits that hold when the work gets hard.

The declared-versus-practiced split has fifty years of theoretical grounding. Argyris and Schön (1974, 1978) introduced espoused theory and theory-in-use as paired constructs at the individual and organizational learning level; Meyer and Rowan (1977) and the institutional-decoupling literature it generated read the same shape at the firm-environment boundary. VCAF operationalizes the distinction at virtue resolution, with declared-baseline-aware metrics the antecedent literature does not attempt.

Sull, Sull, and Chamberlain (2020) studied 562 large U.S. firms with official values statements and found no correlation between published values and how well employees said the firm lived them; on four of nine tracked values (collaboration, customer orientation, execution, diversity), the correlation was negative. VCAF measures that gap directly.

In one operator role I inherited a sales team that had drifted into order-taking. Everyone wore the sales title; no one was running a sales motion. I met the team. I built relationships with them. And I came to know that they had more in them than what they were able to call forth at that point. I was hired in April; six months in, attainment moved from 65% to 120% of quota. The culture improvement drove it. Inside eighteen months, revamped lead distribution, a new approach to customer relationships, a rebuilt comp plan, and the rest of the system changes landed, and the bar moved up from there. The work had finally caught up to the job title.

Aristotle named this in the Nicomachean Ethics: "Moral virtue comes about as a result of habit, whence also its name (ἠθική) is one that is formed by a slight variation from the word ἔθος (habit). From this it is also plain that none of the moral virtues arises in us by nature; for nothing that exists by nature can form a habit contrary to its nature." VCAF closes that gap. The mechanism connecting alignment to operating profit is in the next section; the broader literature lineage is in Appendix G.

---

Why Culture Costs Show Up in Operating Profit

When a firm's declared virtues match its practiced behavior, people act on shared assumptions. They hand off work without re-checking whether the standard still holds. Widen the gap and the cost appears: decisions that should be settled get relitigated, managers second-guess whether leadership meant what it said, and teams default to self-protection rather than coordination. Williamson (1975, 1985) names this dynamic transaction cost; Becker and Murphy (1992) sharpen it as the coordination overhead that limits specialization. The cost surfaces as disengagement (people present but not performing), rework, scrap, decision latency, and the approval gates that multiply when teams stop trusting the information they receive. Edmondson (1999, 2019) traces one channel directly: when a firm declares accountability but punishes early risk-flagging, information that should travel upward early travels late or not at all. Zaheer, McEvily, and Perrone (1998) and McEvily, Perrone, and Zaheer (2003) supply the organizing frame: trust is the asset, alignment is the cultivation mechanism, coordination cost is the unit in which depletion is paid.

Adjusted EBITDA is the dependent variable because that is where the friction lands. Revenue is mostly sales- and marketing-facing and reflects market conditions the framework does not touch. GAAP net income is contaminated by accounting choices. Adjusted EBITDA captures friction from anywhere in the organization: procurement, yield, scrap, collections, hiring cycles, overhead absorption. The dollar that does not arrive at the operating profit line because of a coordination failure is the dollar VCAF is built to find.

Safety and regulatory integrity show the mechanism at its sharpest. A recordable injury triggers a workers' compensation claim, an Experience Modification Rating increase that persists for three years, OSHA exposure, and production downtime. A failed audit triggers penalties and a re-qualification cycle. Both hit the P&L months before they hit revenue. They happen when the organization stops practicing what it declared: the safety standard written on the wall is not the standard held on the floor. Because these failures hit profit faster and harder than any other virtue gap, the framework does not allow strong performance on other virtues to offset them; that is the non-compensatory rule in Step 1.

Six steps run the measurement and the intervention.

---

The Virtue-Culture Alignment Framework

VCAF prices the mismatch between declared and practiced virtue: between the values on the wall and the values in the hallway. The framework applies to any operating business where culture affects realized profit, not only manufacturing; Arete is a specialty manufacturer because that is the worked-example context the author knows best. The dependent variable is Adjusted EBITDA, for the reasons named above; a cross-functional profit measure supports the universality claim. VCAF produces three outputs: a per-virtue alignment measure, an aggregate alignment score, and a directional dollar range for the EBITDA exposed to that mismatch. Figure 1 shows the flow.

Figure 1. VCAF Framework Flow

Declared → Practiced → Metrics → EAI → Diagnose → Map → Intervene → Recheck

Notation in Appendix A; full math in Appendix B; methodology detail in Appendix C.

Step 1: Define Core Virtues and Assign Weights

Leadership identifies three to five strategically material virtues. Each gets a one-sentence definition, behavioral anchors, and a stakeholder weight (wₖ) summing to 1.0. Leadership chooses how aggressively to press each virtue at this step. The outside rater's Dₖ tracks how hard leadership presses through policy, resources, and modeling; a firm that presses hard and only partially delivers shows a larger gap than one that presses lightly and meets its own standard. Certain virtues (safety, regulatory integrity) are inherently non-compensatory: if any virtue's MIₖ exceeds 0.30, intervention triggers regardless of composite score and clears only after two consecutive remeasurements come in below 0.25. Strong performance on fairness cannot offset a failing safety culture; the reason for that asymmetry is in Step 5.

Step 2: Score Each Virtue on Two Dimensions

Each virtue is scored 0 to 100 on two axes. The declared score (Dₖ) captures how strongly leadership reinforces the value through policy, resource allocation, internal and public communication, and personal modeling. Dₖ is scored by a party structurally independent of the leader being assessed (an outside advisor, board member, or rotating internal rater) to avoid self-assessment bias. The practiced score (Pₖ) captures how employees experience the value in daily work.

Step 3: Run the Misalignment Index Assessment

The practiced score blends two data streams: short employee narratives (per-virtue volume and role-tailored prompts specified in Appendix C) scored by a human rater and an AI model against a four-dimension rubric (detail, behavioral specificity, stakes, performance under pressure), and a Likert-scale survey on perceived consistency. Each dimension asks a specific question: did the story name a concrete moment, did it describe actions rather than feelings, was something real at risk, and did the behavior hold when the work got hard. The two streams are normalized and blended 60/40 in favor of narrative because story-grade evidence carries the behavioral specificity the survey cannot. The rubric excludes moral vocabulary as a scoring dimension to avoid penalizing shop-floor workers who describe consequential behavior in plain language. Anonymity is preserved through third-party intake, items rotate cycle to cycle, and scored narratives are triangulated against observable outcomes (turnover, safety incidents, promotion ratios) before any Pₖ score is reported. Worked narratives, dual-rater reliability targets, and the Pₖ computation are in Appendix C.

Step 4: Calculate Alignment Metrics

VCARₖ = Pₖ / Dₖ expresses how much declared intention is actually lived (a ratio measure). MIₖ = |Dₖ − Pₖ| / 100 expresses the absolute integrity gap on a 0 to 1 scale (a gap measure). Two firms with identical absolute gaps but different declared baselines should not score identically; the ratio measure captures what the magnitude measure cannot.

The composite blends both readings. Weighted alignment A = Σₖ (wₖ × VCARₖ^adj) rewards realized practice as a share of declared intention. Weighted misalignment M = Σₖ (wₖ × MIₖ) penalizes the absolute gap. The Ethical Alignment Index combines them with a volatility penalty for misalignment concentrated in one virtue: EAI = λA + (1 − λ)(1 − M) − ρσ.

Parameter ranges for λ and ρ, the interpretation thresholds, the cap on VCARₖ in aggregation, the treatment of latent virtue (behavior that outruns declaration), and the rule that locks λ, ρ, weights, and outcome variables before any score is computed (Appendix E.6) are in Appendix B.

Step 5: Translate the Gap into EBITDA-Risk Terms

The mechanism connecting alignment to operating profit is in the preceding section; this step does the arithmetic. Nonlinear extensions and the SDE alternative for smaller firms are in Appendix B.

EOV = EAI × Productivity Potential, where Productivity Potential is the Adjusted EBITDA the firm would produce at EAI = 1.0. Three principled estimation methods exist (Appendix B). Operators with existing performance data can apply them directly; firms without that data, including those in a leadership transition or standing up a new initiative, fall back to the back-solve described below. Under the fallback, PP = Current Adjusted EBITDA / EAI, which makes EOV = EAI × (Current EBITDA / EAI) simplify to Current Adjusted EBITDA. The restatement is a deterministic translation of EAI into dollar terms, not an independent measurement; it still forces the gap into the units the operator manages. The back-solve forces the operator to commit to weights and a multiple before any score is computed, which makes the framework's underwriting discipline visible before deployment. The back-solve admits a Bayesian reframing (Appendix B, Bayesian Translation of EBITDA at Risk) that integrates a peer-margin prior, recovering credible intervals that scale with measurement reliability. Migration to elasticity-modeled estimates becomes available once a validation cohort accumulates (Appendix E).

EBITDA at Risk is Productivity Potential minus current Adjusted EBITDA. Where profit data is still being validated, revenue serves as an interim proxy that understates the friction profit would reveal. The transaction-context translation (EBITDA at Risk multiplied by an acquisition multiple) applies at any change of ownership; the worked example in Appendix D illustrates the deployment-context calculation. Adjusted EBITDA build-up rules in Appendix B.

Step 6: Build the Contradiction Map and Targeted Intervention

Priority targets are high-weight virtues with the largest gaps; the intervention runs at the management systems that produce the gap. If a firm declares accountability but promotes on politics, the system is the problem and accountability training will not fix it. Structural changes outperform procedural changes (new meeting templates, RACI charts), which outperform communication-only changes (values training, all-hands speeches).

Each data stream does different work. The Likert-scale survey ranks urgency across virtues. Story narratives reveal the mechanism producing the gap, and the firm's operating metrics (yield, turnover, accident rates, on-time delivery) confirm whether an intervention moved practice. Move 4 of Appendix F.1 pre-commits to all three before the change is announced.

The default is P-side action: raise practice to meet declaration. Lowering Dₖ in response to gap pressure converts the framework into alignment theater (caveats in Appendix B).

The Likert × story routing is what makes the intervention plan function-specific. The Likert-scale survey locates urgency at virtue resolution: which virtue is in worst alignment. Stories locate targeting at function resolution: which management system is producing the gap. At Arete, Accountability stories show that public mistake-owning happens after the fact, while the lowest-scoring Likert items cluster on raising forecast risks without penalty and on the review-cadence signal. The combination points at review cadence and forecast-risk surfacing as the contradictions. VCAF's response to Sull et al.'s (2020) finding that values statements do not predict employee experience is to use the narratives themselves to identify management-system contradictions and route intervention there (Appendix F.1, Table D.5b).

Misalignment usually traces to incentives and norms, not individuals. VCAF is a structured decision aid, not a substitute for operator judgment; the gap between the two is itself a diagnostic signal. The full intervention sequence, including three-level metric anchors (Likert, operating metric, story signal), is in Appendix F.1.

---

Illustrative Example: Arete Precision Manufacturing

Arete Precision Manufacturing is a composite mid-market industrial-components company (~60 employees, $12M revenue, $1.8M Adjusted EBITDA at 15% margin) with three core virtues: safety, accountability, fairness. VCAF yields a weighted EAI of 0.74 (Appendix D), with A = 0.72 (proportional alignment) and 1 − M = 0.77 (gap minimization) reported alongside. Accountability is the most misaligned virtue. Back-solving Productivity Potential at $2.43M implies roughly $630K of EBITDA at risk on the $1.8M base, split across virtues by weighted misalignment: ~$290K Safety, ~$215K Accountability, ~$125K Fairness. The point estimates above carry a scoring sensitivity range sized in Table D.4. Transaction-context translation, full arithmetic, the mechanism-to-dollar walk-out, and the per-virtue routing of the intervention plan are in Appendix D (Tables D.1 through D.5b).

---

Limitations, boundary conditions, and falsifiability tests are in Appendix E.

---

Deployment in Practice

The workforce has no reason to trust a leader's diagnostic on day one of any new initiative, whether the leader is new to the firm or opening a new cycle of self-examination. Deployment runs in two phases. Days 1 to 90 are listening and structuring: story collection, observational data (turnover, safety, promotion history), Adjusted EBITDA build-up, and declared-values framework finalization with an outside Dₖ rater. Operational stewardship continues in parallel; VCAF-specific interventions wait for grounded scores. Days 90 to 180 are baseline and commit: a full VCAF run on Adjusted EBITDA, results shared as a commitment rather than a report card, then a quarterly cadence with leadership held to the same scores it asked of the team. Runbook and cadence are in Appendix F.

The first leader who runs this protocol opens cohort n=1, in any firm where the operator owns the gap, whether through ownership, executive role, or a partnership that gives the operator real authority over the management systems VCAF targets. Where the firm carries a declared mission and a workforce large enough to generate the story volume Appendix E specifies, the protocol runs in its full quantitative form; smaller teams run a narrative-only adaptation of the same architecture without the quantitative claim (Appendix F.0). A diagnostic that nobody acts on is expensive theater; the protocol is built to be acted on, in firms where the values on the wall and the values in the hallway are not yet the same words, and the work is to make them the same.

---

Technical Appendix

Supporting Material for Values-Based Operational Decision-Making

---

Appendix A: Notation and Symbols

This appendix catalogs every variable, symbol, and abbreviation used in the framework. Subscript k indexes a single virtue (k = 1, 2, ..., n, where n is the number of virtues being measured, typically 3 to 5).

Scoring inputs (per virtue k)

Per-virtue metrics

Aggregate metrics

Financial translation

Validation statistics

Framework and institutional abbreviations

---

Appendix B: Full Math Derivations

Virtue-Culture Alignment Ratio (VCAR)

Equation. VCARₖ = Pₖ / Dₖ, with aggregation form VCARₖ^adj = min(VCARₖ, 1.0).

Purpose. Measures, per virtue, the ratio of practiced to declared behavior. A VCAR of 1.0 means the business lives what it says; below 1.0 means practice trails declaration; above 1.0 means practice outruns declaration.

How to use. Compute one VCARₖ per virtue. Use VCARₖ^adj in the weighted alignment aggregate A. Retain raw VCAR > 1.0 as a latent-virtue flag for qualitative investigation; prompt a Dₖ recalibration in the next cycle if leadership intends to codify what employees are already doing. The cap is directional, not symmetric: Pₖ below Dₖ remains fully visible in aggregation (the misalignment the framework exists to price), while Pₖ above Dₖ is surfaced qualitatively rather than rewarded numerically. The cap prevents one latent virtue from masking declared-but-unpracticed gaps elsewhere.

Worked walk-out: VCARₖ > 1.0 ("practice outruns declaration"). The case confuses readers on first contact. Mathematically it is straightforward: the practiced score Pₖ exceeds the declared score Dₖ. A worked illustration shows what produces it and what to do.

Take a small welding shop where the founder, now retired, never wrote down a safety doctrine. Dₖ for safety is scored at 70 by the outside rater: there is no posted policy, no scheduled safety briefings, no documented training plan, the values handout makes one passing reference. Pₖ for safety scores 82: stories show operators routinely shutting down a job when a fitting looks wrong; the floor stops a rush order rather than weld a joint that sits at the edge of the tolerance spec; the lead hand has called out the owner's son for cutting a corner and the call held. Three positive narratives in one cycle, low Likert variance, observable safety-incident rate at one-third the industry norm. Raw VCARₖ = 82 / 70 = 1.17.

What this looks like inside the firm: the behavior is real, the language is missing. Floor culture is carrying the safety doctrine that policy and training have not codified. Two risks follow. First, the carrier is fragile. If the lead hand retires, the next-day Pₖ collapses because the practice was never written, taught, or institutionalized. Second, in a deal context the latent virtue is invisible to an incoming leader reading only the values document, which means the firm's represented profile does not reflect it.

What to do. Three moves, in order. Codify: write the doctrine the floor is already living. Specific actions the senior operator and the lead hand actually take become named procedures in the safety manual; the language now matches the practice. Train: the next cohort learns those procedures from the start, instead of learning them by osmosis from one or two senior operators. Recognize: surface the latent virtue in the next leadership communication so the people carrying it see that the owner saw it. Then re-score Dₖ in the next cycle. Dₖ rises (toward 82, in this example), VCARₖ moves toward 1.0, and the latent virtue has been institutionalized rather than left to chance.

Two failure modes to flag. Lowering Dₖ on paper without changing anything is alignment theater (Step 6); the floor practice is unaffected, but the math now reads aligned. Rewarding raw VCAR > 1.0 numerically would create an incentive to under-declare in order to over-perform, which the cap is there to prevent. The cap removes the numerical reward. The qualitative flag preserves the signal.

Misalignment Index (MI)

Equation. MIₖ = |Dₖ − Pₖ| / 100.

Purpose. Captures the absolute gap between declared and practiced behavior, per virtue, on a 0 to 1 scale. Unlike VCAR, MI is symmetric: a Pₖ that overshoots Dₖ registers the same magnitude as a Pₖ that falls short.

How to use. Compute MIₖ per virtue. Use it for (a) the weighted misalignment aggregate M, (b) the non-compensatory trigger check (any MIₖ > 0.30 on a safety or regulatory virtue forces a qualitative override regardless of EAI), and (c) diagnostic routing in intervention design, where the largest MIₖ indicates the virtue most in need of structural change.

Weighted Alignment (A)

Equation. A = Σₖ (wₖ × VCARₖ^adj), where wₖ is the virtue weight and Σwₖ = 1.

Purpose. Aggregates per-virtue alignment into a single 0 to 1 score, weighted by the relative operational importance of each virtue. A rewards what the firm is getting right.

How to use. Set wₖ in advance of measurement, grounded in an explicit rationale (regulatory exposure, customer-facing intensity, safety-critical footprint). Freeze weights across measurement cycles to preserve period-over-period comparability; revisit them only at defined framework reviews. Report A alongside EAI so readers can see the strength component independent of the penalty components.

Weighted Misalignment (M)

Equation. M = Σₖ (wₖ × MIₖ), using the same wₖ as A.

Purpose. Aggregates per-virtue misalignment into a single 0 to 1 score. M prices what the firm is getting wrong.

How to use. Enters EAI as (1 − M) so that lower misalignment raises the score. Because A and M draw on the same weights, the EAI formulation is internally consistent: the virtues that carry the most upside also carry the most downside.

Misalignment Volatility (σ)

Equation. σ = √(Σₖ wₖ × (MIₖ − M)²), the weighted standard deviation of MIₖ across the virtue set. σ is computed from the MIₖ scores at each measurement cycle; it is not a chosen parameter.

Purpose. Measures concentration risk. Two firms can share identical M values while one has its misalignment spread evenly across virtues and the other has it clustered in a single virtue. σ distinguishes the two cases.

How to use. σ is a small distributional adjustment to EAI, on the order of 0.005 at the Arete worked example and capped near 0.10 in the worst plausible case. Most of the EAI signal comes from the A and (1 − M) terms; σ refines the picture between the catastrophic-concentration case (which the non-compensatory trigger handles directly at Step 1) and the case of broad uniform drift (where σ goes to zero). High σ at moderate M points to one virtue carrying the misalignment load and warrants priority intervention; low σ at high M points to broad cultural drift requiring systemic rather than targeted response. σ measures concentration risk, not total risk.

Ethical Alignment Index (EAI)

Equation. EAI = λ × A + (1 − λ) × (1 − M) − ρ × σ, with λ ∈ [0.50, 0.60] and ρ ∈ [0.10, 0.20].

Purpose. The composite operational score. EAI combines the alignment reward (A), the misalignment penalty (1 − M), and a volatility adjustment (ρ × σ) into a single 0 to 1 index that flows directly into the financial translation.

How to use. Fix λ and ρ by policy choice before measurement and lock them through pre-registration (Appendix E.6) so the parameters are framework constants, not tuning dials. λ sets the balance between rewarding what the firm is getting right (higher λ) and penalizing the gap (lower λ); ρ scales the concentration-risk penalty. Publish both parameters alongside EAI so a reader can reproduce the index. Run the λ-ρ sensitivity pair in Appendix D Table D.3 before any go/no-go decision rests on EAI alone. Base case for Arete and the worked example: λ = 0.55, ρ = 0.15.

Choosing λ inside the policy range. Higher λ (closer to 0.60) rewards realized alignment more heavily and is appropriate for firms or sectors where what the firm is getting right is the load-bearing operating asset (knowledge work, professional services, IP-heavy businesses). Lower λ (closer to 0.50) weights gap minimization more heavily and is appropriate for firms or sectors where misalignment surfaces in P&L line items quickly (regulated, safety-critical, or quality-sensitive operations). The base case (λ = 0.55) splits the difference and is the framework default.

Choosing ρ inside the policy range. Higher ρ (closer to 0.20) penalizes concentration risk more heavily and is appropriate for regulated or safety-critical firms where one-virtue clustering of misalignment can produce a single line-item failure (a recordable injury, an environmental citation, a recall). Lower ρ (closer to 0.10) is appropriate where misalignment is broadly distributed and no single virtue carries catastrophic operational risk. The base case (ρ = 0.15) sits at the middle and is appropriate for the typical mid-market specialty-industrial firm.

EAI Interpretation Thresholds

Equation. Qualitative bands over EAI: strong alignment (EAI > 0.85); partial alignment (0.70 ≤ EAI ≤ 0.85); significant drift (EAI < 0.70).

Purpose. Translates the continuous index into a decision vocabulary an operator or board can act on. Thresholds are calibration targets for longitudinal validation, not fixed cutoffs.

How to use. Use the bands as a triage layer, not a substitute for component-level review. A firm at EAI = 0.72 with low σ and high A is a different case from EAI = 0.72 driven by low M and high σ. Always read the band alongside A, M, and σ.

Theoretical Foundations: Coordination Friction and Operating Profit

The body section "Why Culture Costs Show Up in Operating Profit" summarizes the mechanism and names the literature anchors. This appendix develops each tradition in full, supplying the microfoundations for the alignment-to-profit channel.

Internal coordination costs. Williamson (1975, 1985) explains why firms exist at all: they take on the coordination costs markets cannot. Two behavioral assumptions ground the argument: bounded rationality, which limits what can be specified in advance, and opportunism, which makes contracts incomplete. Inside a firm, both assumptions still apply; the governance structure changes. When a firm's declared virtues track its practiced behavior, employees process signals consistently across departments and the implicit cost of cross-functional coordination is low. A wide gap turns every cross-functional decision into a re-negotiation over which value system is binding in this instance, which manager's interpretation governs, and which past commitment was real. Becker and Murphy (1992) sharpen the point with the productivity bridge: the limit on specialization is not market extent but coordination cost and the supply of general knowledge. A firm with high MIₖ on accountability cannot sustain the depth of specialization that high alignment would permit, because every handoff requires verification rather than trust. The friction surfaces in handoff delays, rework, scrap, decision latency, and approval-gate proliferation, all of which hit Adjusted EBITDA on a faster cycle than they hit revenue.

Psychological safety. Edmondson (1999, 2019) provides the empirical grounding for one of the channels through which misalignment depletes coordination capacity. Her finding is that work teams whose members feel safe to raise concerns, admit errors, and surface bad news early outperform teams that do not, and that team-level psychological safety tracks leader behavior and organizational signaling. A firm that declares a virtue like accountability and practices it through penalty for early risk-flagging teaches its workforce that the declared virtue does not hold; the system selects against the behavior the leadership claims to want. Information that should travel upward early travels late or not at all. Disengagement follows. When employees consistently see the firm's declared virtues outranked by what its systems reward, discretionary effort declines; Gallup's Q12 program tracks this dynamic at firm-engagement resolution (Gallup 2024). The framework's Step 6 line ("if you declare accountability but promote on politics, the system is the problem") is Edmondson's mechanism stated in operator vocabulary. The Pₖ measurement layer captures the consequence at virtue resolution; the EAI translation prices it.

Trust as organizing principle. Zaheer, McEvily, and Perrone (1998) showed across 107 buyer-supplier interfirm relationships that trust reduces negotiation costs and improves exchange performance, with interpersonal and interorganizational trust operating as related but distinct constructs. McEvily, Perrone, and Zaheer (2003) generalized the framework to within-firm coordination, treating trust as an organizing principle that shapes how the firm structures itself and how it mobilizes its members. VCAF's Dₖ/Pₖ split measures the within-firm version of the same construct: employees who observe consistency between declared and practiced virtue replenish trust, and coordination overhead falls; a visible gap depletes trust, and the firm pays the cost in more meetings, more documentation, more approval gates, and more internal legal review. Trust is the asset; alignment is the cultivation mechanism; coordination cost is the unit in which depletion is paid; Adjusted EBITDA is where it is realized.

Synthesis. The traditions converge on the same claim. Williamson names it transaction cost; Edmondson reads it as a failure of psychological safety; Zaheer, McEvily, and Perrone treat it as trust depletion. VCAF operationalizes the gap that produces the cost at virtue resolution and translates the result into the units the operator manages. The financial-translation layer rests on this stack: high alignment reduces coordination cost through channels the literature has named but rarely measured at the virtue-and-firm level VCAF specifies.

Ethical Output Value (EOV)

Equation. EOV = EAI × Productivity Potential (PP).

Purpose. The financial translation step. EOV estimates the Adjusted EBITDA a firm should produce at its current alignment level, given its productive capacity (PP).

How to use. Compare EOV to current Adjusted EBITDA to size the operational upside from closing the alignment gap. The difference (PP − Current Adjusted EBITDA) is the basis for Enterprise Value at Risk. Read EOV as an operational estimate, not a valuation: it does not incorporate discount rate, terminal value, or capital structure. Adjusted EBITDA is the right base because regulatory and quality failures hit the P&L long before they hit the top line: a failed environmental or quality audit shows up in penalties, remediation costs, and customer-imposed re-qualification cycles long before it shows up in lost orders, and a single recordable injury adds workers' compensation claims and OSHA citation exposure on top of the experience modification rate noted in Step 5. Revenue reflects market conditions, pricing power, and demand cycles in addition to coordination quality, so the alignment signal arrives in revenue at a longer lag than it does in operating profit; GAAP net income is contaminated by accounting choices and one-time items. Adjusted EBITDA exposes the friction those bases hide. For smaller firms (typically below ~$1M EBITDA) where Adjusted EBITDA is unstable due to small-business owner-economics commingling, Seller's Discretionary Earnings (SDE) is the analogous base; the same VCAF math applies with SDE substituted for Adjusted EBITDA throughout.

Productivity Potential (PP)

Equation. PP admits three principled estimation methods plus a default fallback.

• (1) Internal historical: PP = Adjusted EBITDA_{t−n} / EAI_{t−n}, where t−n is a prior period with measured EAI.
• (2) External benchmark: PP = (peer best-in-class Adjusted EBITDA margin) × (current revenue), drawn from a defined comparable set.
• (3) Elasticity model: PP derived from a cohort regression ΔAdjusted EBITDA = β × ΔEAI.
• Fallback: PP = Current Adjusted EBITDA / Current EAI.

Purpose. PP estimates the firm's attainable Adjusted EBITDA at EAI = 1.0 as the company stands today. It is the productive ceiling against which current performance is measured. PP recomputes when the firm's structural capabilities change (capacity additions, capital deployment, customer-mix restructuring, scope changes); the framework treats PP as a snapshot under a defined firm-level composition, not as a fixed property of the firm.

How to use. Method selection depends on data availability. Method (1) requires longitudinal EAI on the same firm. Method (2) requires virtue-resolution peer data, which is rarely available for private firms. Method (3) requires a multi-firm cohort. The fallback produces no independent estimate (it is mathematically equivalent to method (1) at t = 0 with no comparison period) and functions as an deployment-time placeholder until cohort data supports method (3). Migration to method (3) is committed in the validation plan (Appendix E). Report the method used alongside every PP figure.

Bayesian Translation of EBITDA at Risk

The default fallback PP estimation method (PP = Current Adjusted EBITDA / Current EAI) carries a known circularity: it produces no independent estimate of productivity potential, only a restatement of current operating profit in EAI-deflated terms. The framework discloses the limitation in Step 5 and treats the back-solve as a deployment-time diagnostic rather than as measurement. This appendix names the principled extension that dissolves the circularity once outside data is integrated.

The circularity, plainly. The back-solve computes PP using Current EBITDA and Current EAI. EBITDA at Risk is then PP minus Current EBITDA. Both numbers come from the same two inputs, which means the dollar figure carries no information beyond what those inputs already contain. The deterministic point estimate appears more confident than the inputs warrant.

The principled fix. The circularity dissolves when an independent estimate of PP is integrated from outside the firm's own current numbers. Peer-margin data supplies that estimate: top-quartile EBITDA margins for the firm's industry and scale (RMA Annual Statement Studies, IBISWorld industry reports, IBBA Market Pulse) give a defensible range for what the firm should produce at full alignment. Combining the peer estimate with the firm's measured EAI and the measurement uncertainty around it produces a range for EBITDA at Risk rather than a single point. The range scales with input quality. A deployment that fails reliability gates produces a wider range an operator can read as "the instrument is not yet fit for translation"; a deployment that clears reliability gates produces a tighter range that earns the operator's confidence. The back-solve is the limiting case of the fuller treatment when no peer data is available and measurement uncertainty is set to zero.

Why this matters. The principled extension addresses the methods-review attack that the back-solve dollar figure is a tautology. The full probabilistic treatment is forthcoming work; it requires calibration against the reliability statistics in Appendix E.4. The framework discloses both the current limitation and the principled path past it. The Pₖ ±10 sensitivity range in Appendix D Table D.4 already approximates what the full extension will produce for the Arete worked example; the sensitivity table and the formal credible-interval analysis express the same uncertainty in different forms.

Enterprise Value at Risk

Equation. EV at Risk = (PP − Current Adjusted EBITDA) × Acquisition Multiple.

Purpose. Sizes the valuation consequence of the alignment gap by applying the prevailing transaction multiple to the EBITDA shortfall.

How to use. Use the multiple appropriate to the transaction context (acquirer's target multiple, peer-set average, or held-asset comparable). Report EV at Risk as a range bracketing the multiple rather than a point estimate. Treat it as a diligence signal and intervention-sizing tool, not a price bid.

Temporal scope. The acquisition multiple is a moment-in-time market parameter, not a firm-level property. EV at Risk is therefore a snapshot under a specific market-and-firm composition, not a forecast that survives the intervention it sizes. It is operationally meaningful at the moments when the firm is being valued against an outside transaction: a change of ownership and the exit. Between those moments the operating quantity is EBITDA at Risk, which the operator can read continuously across the deployment period; see Appendix D for the temporal-frame deployment.

Nonlinear Extensions

Equation. Two specifications are flagged for future calibration.

• Sigmoidal: EOV = PP / (1 + exp(−γ(EAI − EAI₀))), where γ is a steepness parameter and EAI₀ is the inflection point. EAI enters EBITDA through an S-shaped curve rather than a straight line, imposing threshold effects.
• Piecewise-linear: different slopes above and below one or more breakpoints, to reflect regime shifts.

Why a nonlinear form is needed at all. The linear form treats every point of EAI as worth the same dollars in EBITDA, regardless of where the firm sits on the alignment scale. Operators who have run cultural turnarounds describe a different shape: months of work with no visible movement, then sudden traction once a critical mass forms, then diminishing returns once the friction has been squeezed out. The linear form cannot represent that experience. Two specifications match it more honestly.

The sigmoidal form addresses threshold dynamics. Below a floor of alignment the team cannot capture coordination-driven gains (a workforce that does not trust each other will not coordinate yield improvements, even with strong declared values). Above a ceiling there is no friction left to extract (a high-alignment firm has already reclaimed most of the operating slack). Linear math sees neither edge; the sigmoid does.

The piecewise-linear form addresses regime shifts. Structural events (comp-plan rebuilds, leadership changes, M&A integration, regulatory step-changes) move the firm from one alignment-to-profit relationship to another. The same firm before and after a comp-plan rebuild has two different slopes, and a single linear coefficient averages across regimes that should be modeled separately.

What each form makes visible that linear cannot. The sigmoidal form locates the inflection point: the EAI level at which alignment work starts paying back at full rate. For an operator, that is the answer to "how much culture work do I need to do before I see profit movement?" The linear form has no such threshold and gives no early-warning signal that the firm is below the productive zone. The piecewise-linear form locates the structural breakpoints: the EAI ranges where the slope changes, with each segment's slope reportable separately. For an operator, that is the answer to "did this comp-plan rebuild change the rules of how alignment translates to dollars?" The linear form averages across regimes and hides the regime change.

Why the sigmoid is the leading candidate. Operator-experience evidence (turnarounds feel nonlinear with thresholds) plus a structural argument (coordination failure is a tipping-point phenomenon: below the floor every department fights friction, above the inflection friction drops fast). The piecewise-linear form is easier to calibrate once data exists and is the right choice for firms with identifiable regime-change events; the sigmoid is the right choice for firms whose alignment trajectory is gradual and uninterrupted.

Directional bias of the linear simplification. The linear form is a deliberate first-order approximation, retained for interpretability and because nonlinear forms require longitudinal data to calibrate. While that approximation is in use, two biases are predictable. For low-EAI firms (below the sigmoid's inflection), the linear form underestimates EBITDA at Risk: alignment work near the floor produces less profit movement than the linear math implies, but the same firm has more dollars at stake once it crosses the inflection. For high-EAI firms (above the sigmoid's ceiling), the linear form overestimates EBITDA at Risk: incremental alignment produces less profit movement than the linear math implies, because most of the coordination friction has already been reclaimed. Operators reading their own VCAF output should weight the directional range accordingly. Naming this bias on the page pre-empts the reviewer challenge that the linear form has a known systematic error.

How to use. Retain the linear form for the initial framework on interpretability grounds. Pre-register the form per Appendix E.6. Test sigmoidal and piecewise-linear specifications in the validation plan (Appendix E) against longitudinal data; adopt whichever specification fits best once the cohort supports estimation. The choice between sigmoidal and piecewise-linear is itself an empirical question the cohort answers.

A separate question the cohort can address is whether MIₖ should retain sign rather than magnitude. The current form, MIₖ = |Dₖ − Pₖ| / 100, treats over-performance and under-performance symmetrically; a signed form would let the operator distinguish the two. The framework holds the symmetric form because no cohort data yet exists to test whether direction carries operational signal a magnitude form misses; a directional version becomes a candidate for revision once the data warrants.

Boundary Conditions

The framework operates across industries because its central construct, the gap between declared and practiced virtue, is a universal organizational phenomenon. Some structural conditions support it; others weaken or break it. This appendix names both sets explicitly so a leader deploying VCAF reads the score against the right structural backdrop.

Where VCAF is strongest. The framework reads most cleanly in firms where coordination is a primary determinant of operating performance:

• High interdependence in the work. When task completion requires sustained collaboration across functions and the cost of a coordination failure is paid in P&L line items (rework, downtime, scrap, missed deliveries, regulatory exposure), virtue-culture alignment translates most directly into operating performance. Examples cross industry lines: manufacturing with cross-shift handoffs, professional services with team-based delivery, healthcare with care-team handoffs, publishing with editorial-to-production handoffs, software with long approval cycles. Approval-by-committee is the institutional form of this friction across firm sizes and industries, and is one of the largest measurable taxes on productivity VCAF surfaces.
• Owner-operator transitions. When the operator has the standing to act on the diagnosis and the time horizon to see intervention through, the framework's intervention layer translates from diagnostic to financial movement.
• Regulated or safety-critical environments. Industries where misalignment hits P&L line items (workers' compensation, OSHA, regulatory recalls, malpractice claims) on a faster cycle than it hits revenue allow the EAI-to-EBITDA channel to be observed during the operator's intervention window.
• Stable industry conditions. When the dominant variation in firm-level performance is internal rather than market-driven, VCAF's coordination-cost channel is detectable above the noise of demand shocks and competitive disruption.

Where VCAF will fail or weaken. Settings where the alignment-to-EBITDA chain breaks down:

• Atomized work with weak coordination requirements. In commission-based sales organizations, contractor-heavy services, or any setting where individual incentives dominate and cross-functional coordination is minimal, the alignment construct still holds, but the operating-profit channel is weak because there is little coordination cost to reduce.
• Pure-rent or capital-structure-dominated businesses. Real estate holding, IP portfolios, financial arbitrage, and similar models produce operating profit through capital structure and market timing rather than through the coordination of work. The framework's diagnostic still applies; the financial translation does not.
• Firms with strong opposing subcultures. Culture permeates downward through hierarchy in firms of any size, and firm-level alignment readings work for most deployments. Some firms harbor subcultures (functional, geographic, business-unit, or generational) holding values that diverge sharply from the dominant declared values. VCAF run at the firm level then averages over heterogeneity that segment-level measurement would expose. Deployment in such firms requires running the framework at segment resolution.
• Firms in acute distress. During acute liquidity, regulatory, or going-concern crises, coordination friction does not disappear, but the operator's attention is committed to survival decisions. VCAF deployment waits; the listening phase remains appropriate, but the intervention layer holds until the firm is past the immediate threat.
• Industries undergoing structural disruption. When demand-side and competitive shocks dominate firm-level performance variation, the alignment-to-EBITDA channel is overwhelmed by exogenous variation the framework does not model.

These structural conditions are assessment inputs before taking on a firm and deployment caveats after. A leader who screens an opportunity against them before running VCAF avoids a misread of the alignment-to-EBITDA chain in either direction.

---

Appendix C: Misalignment Index Assessment Methodology

The assessment architecture (structured elicitation scored against a fixed rubric by independent raters, with inter-rater reliability as a published validation target) follows the methodological precedent established by Bloom and Van Reenen's (2007) original 732-firm survey across four countries; the World Management Survey program has since extended that methodology to more than 20,000 management interviews across 35 countries (worldmanagementsurvey.org). VCAF adopts that elicitation pattern and extends it in three ways. The unit of measurement shifts from operational management practices to virtue-level behavior. Scoring is dual-rater (human plus frontier-model LLM) rather than human-only. The instrument outputs a per-virtue alignment metric tied directly to a financial translation rather than a single management-quality index.

C.1 Story Elicitation

Employees receive a per-virtue prompt framed as an invitation to recall a specific moment. Example prompts used in a pilot cycle:

• "Tell me about a time in the last twelve months when someone on this team made a decision that put safety ahead of schedule or cost."
• "Describe a recent moment when a peer or manager owned a mistake in front of the group."
• "Walk me through a promotion or pay decision you watched happen that you thought was clearly fair, or clearly not."

Prompts are open-ended and concrete; they ask for stories rather than opinions. Responses are collected anonymously through a third-party intake (typed or dictated), with a minimum target of eight to twelve stories per virtue per cycle.

C.2 Story Rubric

Each narrative is scored 0 to 25 on four dimensions, summing to 0 to 100:

1. Detail. Concrete people, places, dates.
2. Behavioral specificity. Named action versus abstraction.
3. Stakes. What was at risk in the moment.
4. Performance under pressure. Whether the behavior held when it was hardest.

Moral or emotional vocabulary is deliberately excluded as a scoring dimension to avoid penalizing shop-floor workers whose accounts are concrete but plain (rationale developed in Step 3 of the body).

C.3 Worked Story Examples

Three representative narratives from the Arete pilot cycle, one per virtue. Scores shown are the rounded average of human and AI raters.

Safety (score: 89). "Tuesday second shift, Line 3, press operator Javier caught the die-change crew cutting corners on lockout-tagout because we were already forty minutes behind for the Ford release. He shut the line down, walked the crew back through the procedure step by step, and told the shift supervisor in front of everyone that the line was not restarting until every lock was verified. We lost another twenty minutes. Nobody argued."

Detail 23 (named shift, line, operator, customer, duration). Behavioral specificity 24 (stopped the line, walked through procedure, spoke to supervisor). Stakes 22 (lockout-tagout on a die change, customer release in play). Performance under pressure 20 (held against schedule pressure).

Accountability (score: 93). "Quality kicked back the first article on the new housing job in October. I was the one who signed off on the fixture and I missed it. I sent a shop-wide email that morning saying I was the source of the defect, what I had missed, and what we were changing on the setup sheet. Two operators told me afterward it was the first time they had seen a production manager do that."

Detail 24 (month, part, role action, specific fix). Behavioral specificity 24 (wrote the email, named the source, changed the setup sheet). Stakes 22 (reputational and operational on a new customer job). Performance under pressure 23 (owned it publicly, in writing, same day).

Fairness (score: 89). "When the cell lead seat opened up in February, everyone on the floor assumed it was going to Dan because he and the plant manager ride bikes together. It went to Maria. She had run the highest yield numbers on the floor for two straight quarters and had trained three of the new hires. The plant manager walked the criteria through at the Monday huddle without being asked."

Detail 22 (month, relationship, named people, tenure of metrics). Behavioral specificity 23 (posted criteria, walked through at huddle). Stakes 22 (a fairness signal read by the whole floor). Performance under pressure 22 (held against the social pull toward the friend).

C.4 Survey Design and Rescaling

Likert items (1 to 5) measure perceived behavioral consistency per virtue. Items rotate across cycles to prevent rehearsed responses. Sample items:

• Accountability: "I can raise forecast risks without fear of penalty."
• Accountability: "Leaders openly discuss mistakes and lessons learned."
• Safety: "When safety and schedule conflict, safety wins on this floor."
• Fairness: "Promotion decisions in this company reflect performance, not politics."

Raw item averages (1 to 5) are rescaled linearly to 0 to 100 using (mean − 1) × 25. Open-ended prompts accompany the Likert block to capture content the scale misses.

C.5 Worked Pₖ Computation (Accountability at Arete)

Eleven usable stories came in for Accountability in the pilot cycle. Per-story rubric totals from the dual-rater average:

58, 64, 55, 71, 62, 48, 67, 53, 60, 69, 54

Story mean: 60.

Likert item mean across the Accountability items: 2.8 on the 1 to 5 scale. Rescaled: (2.8 − 1) × 25 = 45.

Blend: Pₖ = 0.60 × 60 + 0.40 × 45 = 36 + 18 = 54.

This matches the Accountability Pₖ reported in Appendix D. The story mean sitting fifteen points above the Likert rescale is itself a signal: employees can name specific moments of accountability, while daily experience is less consistent than those moments suggest.

C.6 AI Rater

Each story is rated independently by a human and a frontier-model LLM. Model identifier and prompt version are logged with each scoring cycle for traceability. A frontier model run on the same prompt a quarter apart cannot be expected to produce identical outputs; the dual-rater design and the cycle-by-cycle reliability gates in Appendix E.4 are how the framework manages model drift. An example of a full production prompt follows. Anchor language should be calibrated to the firm's virtue definitions and industry vocabulary; the dimension structure, the moral-vocabulary exclusion, and the JSON output contract are framework-level and should not be changed cycle-to-cycle without re-baselining.

SYSTEM

You are a rater for the VCAF Misalignment Index Assessment. Your job is to score a single employee narrative against a fixed 0 to 100 rubric. You are one of two independent raters; your scores are averaged with a human rater and compared via weighted Cohen's κ. Consistency across cycles matters more than generosity.

TASK

Read the narrative. Score it on four dimensions, 0 to 25 each. Return strict JSON with one-sentence justifications per dimension and a total.

DIMENSIONS

1. Detail (0 to 25). Concrete people, places, dates, amounts, durations.

• 0 to 5: no identifying detail; generic ("sometimes people do X").

• 6 to 12: one or two concrete anchors (a name, or a date).

• 13 to 19: multiple concrete anchors (named person, named location, time frame).

• 20 to 25: rich specificity (named people and roles, specific shift or date, named customer or project, quantified stakes).

2. Behavioral specificity (0 to 25). Named actions rather than abstractions.

• 0 to 5: abstractions only ("leadership stepped up").

• 6 to 12: one named action ("she spoke up").

• 13 to 19: sequence of named actions ("she stopped the line and walked through the procedure").

• 20 to 25: named actions tied to named people with observable outcomes.

3. Stakes (0 to 25). What was at risk in the moment (financial, safety, reputational, regulatory, relational).

• 0 to 5: no stakes identifiable.

• 6 to 12: vague stakes ("it mattered").

• 13 to 19: stakes named but not quantified ("a customer release was on the line").

• 20 to 25: stakes named and grounded in operational or financial consequence (specific customer, specific deadline, specific regulatory exposure).

4. Performance under pressure (0 to 25). Whether the behavior held when it was hardest. Default to 0 if no tension is described.

• 0 to 5: no pressure present, or the behavior folded.

• 6 to 12: minor pressure, behavior held.

• 13 to 19: meaningful pressure (schedule, cost, authority gradient, social cost), behavior held.

• 20 to 25: significant pressure (active conflict, visible personal cost, authority pushback), behavior held visibly.

SCORING RULES

• Score only what is present in the narrative. Do not infer unstated virtue.

• Do not reward moral or emotional vocabulary. "Courageous," "integrity," and "did the right thing" score no higher than plain language describing the same action.

• Reward concrete nouns and verbs: named shifts, named lines, named customers, named parts, named dates, named roles.

• If two events appear in one narrative, score the stronger. Do not average across events within a narrative.

• If the narrative describes values-incongruent behavior (a moment the company fell short), score it on how clearly the failure is rendered. A well-rendered failure story receives a high rubric score and is coded as a negative observation at aggregation.

• If the narrative is empty, off-topic, or unreadable, return all zeros and set valence to "non-scorable."

CALIBRATION EXAMPLES

Example A (safety, target total ≈ 89).

Narrative: "Tuesday second shift, Line 3, press operator Javier caught the die-change crew cutting corners on lockout-tagout because we were already forty minutes behind for the Ford release. He shut the line down, walked the crew back through the procedure step by step, and told the shift supervisor in front of everyone that the line was not restarting until every lock was verified. We lost another twenty minutes. Nobody argued."

Targets: detail 23, behavioral_specificity 24, stakes 22, performance_under_pressure 20, total 89, valence positive.

Example B (accountability, target total ≈ 45).

Narrative: "My manager is usually pretty good about admitting when she's wrong. A few weeks ago she made a mistake on the forecast and she said something about it in the team meeting. It felt like she was taking responsibility."

Targets: detail 11, behavioral_specificity 10, stakes 12, performance_under_pressure 12, total 45, valence positive.

Rationale: one named relationship, no named people; one vague action; vague stakes; no visible pressure.

Example C (fairness, target total ≈ 7).

Narrative: "I think our promotion process is pretty fair most of the time. People generally get what they deserve here."

Targets: detail 2, behavioral_specificity 2, stakes 3, performance_under_pressure 0, total 7, valence non-scorable.

Rationale: opinion rather than narrative; no specific event.

OUTPUT CONTRACT

Return strict JSON matching this exact shape. No prose before or after.

{

"detail": <int 0-25>,

"detail_justification": "<one sentence>",

"behavioral_specificity": <int 0-25>,

"behavioral_specificity_justification": "<one sentence>",

"stakes": <int 0-25>,

"stakes_justification": "<one sentence>",

"performance_under_pressure": <int 0-25>,

"performance_under_pressure_justification": "<one sentence>",

"total": <int 0-100>,

"valence": "<positive|negative|non-scorable>",

"virtue": "<safety|accountability|fairness|...>"

}

NARRATIVE

<<<STORY>>>

Human and AI scores are averaged per story. Inter-rater agreement is monitored via weighted Cohen's κ with a target threshold of ≥0.70. Cycles where agreement falls below threshold trigger rubric recalibration before any scores are released.

C.7 Blend Rationale and Gaming Mitigation

Pₖ = 0.60 × qualitative + 0.40 × quantitative. Stories are the highest-resolution behavioral signal the firm produces; the rubric (detail, behavioral specificity, stakes, performance under pressure) makes that signal scoreable. Surveys are retained for reliability and comparability.

Anonymity is enforced through third-party collection. Survey items rotate across cycles. Results are triangulated against observable outcomes (safety incident rates, turnover by department, internal promotion ratios, etc.), so self-reported alignment that contradicts behavioral data is flagged at the point of aggregation rather than at the point of reporting.

The story stream carries a structural defense against KPI gaming that the Likert stream does not. Stories capture behavior; Likert items capture self-reported virtue scores. A workforce that learns the Likert items can drift toward the answers the items reward; a workforce that submits stories about specific events on specific shifts has no equivalent maneuver. The 60/40 blend in favor of narrative is partly an honesty argument and partly a gaming-resistance argument.

C.8 Story Coding for Intervention Targeting

The Likert × story routing in Step 6 turns narrative content into specific function-level intervention candidates. The discipline that keeps that routing honest is laid out here. Four rules govern any intervention candidate read out of the narrative corpus.

Three-narrative convergence. No single story produces an intervention candidate. The threshold for treating a story-derived signal as a routing input is at least three independent narratives in the same cycle pointing at the same management system. A single story is qualitative evidence, not a coded signal.

Both-extremes walk. The function-level coding requires reading both the top-of-virtue and the bottom-of-virtue narratives for the priority virtue. Top-of-virtue stories show what the firm gets right and where the practice already lives in the system; bottom-of-virtue stories show where the contradiction is. Routing the intervention requires both reads.

Candidate-replication standard. A management system is treated as a routing candidate only if it appears in stories submitted by employees in at least two distinct functional or hierarchical positions. A pattern visible only to one function or one hierarchical layer is read as functional information, not as a firm-level routing target.

Privacy disclosure. Routing-grade narratives are coded by management system, not by named individual. Where a story is identifiable in the routing record (a manager named, a sequence of decisions traceable), the rater team reduces the routed entry to the management-system pattern and retains the full text only in the rater archive (Appendix C.6 prompt versioning). The privacy-disclosure rule in F.0 governs the workforce-facing announcement; this rule governs internal coding hygiene.

These rules guard the routing claim against two failure modes: a single vivid story carrying weight no signal at this scale should carry, and a firm-level intervention designed off a single function's view of the misalignment.

---

Appendix D: Arete Worked Example

Arete Precision Manufacturing (composite, illustrative only, not a backtest on a real case): ~60 employees, $12M revenue, $1.8M Adjusted EBITDA (15% margin), specialty industrial components serving automotive and energy clients. Scaled to represent a mid-market specialty manufacturer; the same math applies at larger scale with proportional dollar figures.

Table D.1. Virtues, Weights, Declared & Practiced Scores

Table D.2. VCAR and MI by Virtue

Step 1. Weighted aggregation of VCAR, MI, and σ.

Form:

A = Σₖ (wₖ × VCARₖ^adj)

M = Σₖ (wₖ × MIₖ)

σ = weighted population SD of MIₖ across virtues

Arete:

A = (0.45 × 0.74) + (0.30 × 0.68) + (0.25 × 0.76) = 0.72

M = (0.45 × 0.23) + (0.30 × 0.26) + (0.25 × 0.18) = 0.226

σ = 0.030 (weighted population SD of MIₖ, two-decimal precision). Low volatility; alignment is consistently suboptimal across virtues rather than one catastrophic outlier.

Step 2. Ethical Alignment Index.

Form:

EAI = λA + (1 − λ)(1 − M) − ρσ

Parameters (pre-registered per Appendix E.6): λ = 0.55, ρ = 0.15.

Arete:

EAI = 0.55(0.72) + 0.45(0.774) − 0.15(0.030) = 0.399 + 0.348 − 0.005 = 0.742

Rounded to 0.74 in the main body. Reported alongside: A = 0.72 (proportional alignment), 1 − M = 0.77 (gap minimization). The two readings move together at Arete; firms whose declared baselines diverge from their absolute gaps will see them separate.

Step 3. Ethical Output Value.

Form (default back-solve fallback; the three principled Productivity Potential estimation methods are in Appendix B):

Productivity Potential = Current Adjusted EBITDA / Current EAI

EBITDA at Risk = Productivity Potential − Current Adjusted EBITDA

Enterprise Value at Risk = EBITDA at Risk × Acquisition Multiple

Arete:

Productivity Potential = $1.8M / 0.74 ≈ $2.43M

EBITDA at Risk = $2.43M − $1.8M ≈ $630K

Enterprise Value at Risk = $630K × 4.5× ≈ $2.8M (at 4.0×, ~$2.5M; at 5.0×, ~$3.2M)

Interpretation: Accountability is the priority intervention. The Accountability stories at Arete (story mean 60) show that public mistake-owning already happens; the Likert results (mean 2.8) show daily experience is inconsistent, with the two lowest-scoring items clustered on raising forecast risks without penalty and on the review-cadence signal. The intervention therefore targets review cadence and promotion criteria; values training would miss the mechanism. A structural audit precedes any values-rollout communication.

The $630K EBITDA at Risk is the operator's sizing number: a directional dollar range sized to the alignment gap. It is not a promised recovery. Step 5 names the math as deterministic; the range disciplines pre-commitment to the intervention plan rather than booking expected savings. At the moment of a change of ownership, the same gap can be sized in transaction terms: $630K × 4.5× ≈ $2.8M of enterprise value at risk under current multiples, roughly 35% of an ~$8M enterprise value. The translation anchors to two transaction moments. The entry number expresses the gap in the multiple under which the firm was valued; the exit number expresses it in the multiple at which the firm will be valued next. Between those moments the operating quantity is EBITDA at Risk. Heraclitus held that "you cannot step twice into the same rivers; for fresh waters are ever flowing in upon you" (DK 22 B12; trans. Burnet, 1930). An operator who applies VCAF and closes the alignment gap is running a different firm than the one assessed. The EBITDA has moved, which moves the operating risk, which moves the multiple. Enterprise Value at Risk at one point in time does not forecast value at a later one; it is computed under the prevailing multiple. From the first day of intervention, the EBITDA starts moving and the relevant multiple for valuing the changed firm shifts with it.

Table D.3. Parameter Sensitivity on the Arete Example

EBITDA at Risk is the operating metric; EV at Risk is the transaction-context translation, held at the base-case multiple of 4.5×. The multiple environment is exogenous to VCAF and is not varied in this table or in D.4. The λ range [0.50, 0.60] reflects the policy choice that A and (1 − M) carry roughly equal weight; the ρ range [0.10, 0.20] reflects the policy choice that concentration risk is a real but minority component of the composite. Both parameters are pre-registered (Appendix E.6) and locked across measurement cycles.

Across plausible (λ, ρ) choices, EAI varies by less than 0.005 and EBITDA at Risk by roughly $16K on a $626K base; the corresponding EV at Risk swing is ~$80K on a $2.8M base. A 0.005 swing in EAI is below the precision of the underlying Dₖ and Pₖ scoring. Parameter choice is not where this model's uncertainty lives; the uncertainty lives in the underlying Pₖ measurements (Table D.4). Both parameters are pre-registered per Appendix E.6 and locked across measurement cycles to preserve period-over-period comparability.

Table D.4. Pₖ Measurement Sensitivity

As in D.3, the acquisition multiple is held at 4.5× and the multiple environment is not modeled. Composite parameters held at λ = 0.55, ρ = 0.15.

Note. At Pₖ −10, Safety MIₖ rises to 0.33 and crosses the non-compensatory threshold. This triggers intervention regardless of composite score, as designed: Step 1 makes Safety non-compensatory, so a safety collapse short-circuits the aggregate. Derivation: Safety Pₖ = 65 − 10 = 55; MIₖ = |88 − 55| / 100 = 0.33.

Pₖ measurement is the dominant source of uncertainty in the framework. A ±10-point shift in Pₖ across all virtues moves EBITDA at Risk between roughly $310K and $1.06M, a 3.4× spread; the corresponding EV at Risk under current multiples ranges $1.4M to $4.8M. VCAF outputs a range. The range itself is the output; treating any single number inside it as a hard point estimate misuses the framework. The width of that range scales with how rigorously Pₖ is measured. This is why Phase 1 of the deployment runbook allocates ten weeks to story collection, behavioral triangulation, and dual-rater scoring before any number is reported. The framework's credibility is only as good as the input data, and the model is honest about that.

Table D.5a. Mechanism-to-Dollar Walk-Out: Allocation Arithmetic (illustrative)

The $630K aggregate EBITDA at Risk breaks down by weighted misalignment, virtue by virtue. Each virtue's contribution to M = Σₖ (wₖ × MIₖ) = 0.226 (Step 1 above) sets its share of the gap. Multiplying that share by the aggregate $630K gives an illustrative dollar attribution per virtue. The arithmetic is exact against M; the dollar attribution inherits the back-solved aggregate's status as a deployment-time diagnostic (Step 5; principled extension in Appendix B Bayesian Translation of EBITDA at Risk).

Table D.5b. Mechanism-to-Dollar Walk-Out: Per-Virtue Mechanism and Metric

Each virtue's allocated dollar share routes through one or two candidate mechanisms (the systems the operator walks at F.1 Move 1) to one or two operating metrics (the line items the operator pre-commits to at F.1 Move 4). Library-level mechanism and metric inventories per virtue are in Appendix F.5; the table below is the worked-example slice for Arete.

How to read these tables. The dollar shares in D.5a and the misalignment severities in D.2 answer different questions. An operator focused on the largest EBITDA at Risk chunk would reasonably pick Safety: its $290K share is the largest in the firm. The framework's priority-intervention call goes elsewhere. Safety's $290K share is large because its weight (0.45) is large, not because Safety is the most broken virtue: Safety's MIₖ (0.23) sits below the 0.30 non-compensatory trigger. Safety is well-aligned and dollar-heavy. Accountability's MIₖ (0.26) is the highest in the framework. Accountability is the actively-misaligned virtue and the priority intervention in F.1 even though its dollar share ($215K) is smaller. The Safety dollar share is what the framework asks the operator to protect through preventive intervention; the dollars at stake if Safety drifts into the trigger zone are the largest in the firm. Fairness (MIₖ = 0.18; $125K share) is the smallest by both measures, and the mechanism fix at sub-50 scale is documentation discipline rather than banding redesign; the read-out is the next promotion or compensation decision, two to three quarters out.

The dollar shares are exact-arithmetic decompositions of the back-solved aggregate; they inherit its status as a deployment-time diagnostic from Step 5. They do not claim to be independent dollar estimates, and they do not claim that closing one virtue's gap leaves the others unchanged. In practice the same comp plan often punishes both Accountability and Fairness, so the shares overlap operationally even where they decompose cleanly arithmetically. What the allocation provides is a routed claim from the $630K aggregate to specific mechanisms the operator can walk and specific metrics the next measurement cycle can ratify or refute. F.5 sketches the library; D.5 runs the worked example end to end.

---

Appendix E: Validation Study Plan

Every number VCAF produces, from VCARₖ to EBITDA at Risk, carries a claim about what the instrument is measuring and how reliably it can be compared across time, raters, and firms. The validation plan is how that claim is checked. Reliability tests confirm the instrument is consistent enough for period-over-period and cross-firm comparison. Predictive-validity tests confirm the EAI-to-EBITDA linkage is empirically supported rather than only theoretically argued. Benchmarking against established instruments confirms VCAF does not contradict prior work on their overlap and adds diagnostic resolution where those instruments do not reach. A reader who takes the dollar figures seriously is entitled to see the evidence standards VCAF commits to meeting.

E.0 Limitations, Boundary Conditions, and Falsifiability

Scores are approximations. Dₖ and Pₖ rest on subjective assessments shaped by perception bias, social-desirability effects, and measurement noise. The dollar range VCAF reports widens or narrows in direct proportion to that measurement quality; Table D.4 sizes the implied range, and the Phase 1 reliability gates in E.4 set the standard before any score is published. The instrument is most needed and least reliable in low-psychological-safety environments (Edmondson 1999). Anonymity, dual-rater design, behavioral triangulation, and a small-firm runbook anchored in observable data are the operational response (Appendices C and F).

Construct validity is a real attack surface. A reader could argue VCAF is measuring general management quality under ethical language. The response: virtues do not float above management systems; they become real only through them, and VCAF locates ethical drift inside the systems that reward, punish, signal, and select behavior. Convergent and discriminant validity tests against engagement and management-practice instruments (E.4, E.7) are how the claim is checked.

Treat EAI as a directional index. Organizational-behavior research supports the link between alignment and profit; this paper does not claim to prove causation. Cultural change likely compounds nonlinearly with lags (linear form retained for interpretability; nonlinear alternatives, plus boundary conditions where the framework will fail or weaken, are in Appendix B). The n=1 commitment in E.1 below converts the theoretical link into a testable one, with the operator's capital underwriting the test. Empirical tests of the framework are designed to run longitudinally across a validation cohort, where the same firms are remeasured over multiple intervention cycles; that design is what converts a directional link into a falsifiable one.

Once EAI becomes a board-reported target, Goodhart's Law applies: behaviors optimize for the measure rather than the virtue. Rotating items, behavioral triangulation, and story weighting slow the drift; stories carry an extra defense by capturing behavior rather than self-reported virtue scores, which is why the surveys-and-stories layer resists KPI gaming a pure-Likert instrument cannot. Because the operator who deploys VCAF also owns the EBITDA outcome it estimates, the outside Dₖ rater only partially neutralizes that incentive.

The framework is falsifiable: if within-firm ΔEAI does not correlate in direction with within-firm ΔAdjusted EBITDA across four quarterly cycles, the linear translation fails and VCAF reverts to a diagnostic-only instrument (E.8 also retires the financial translation if Pₖ reliability targets fail across two consecutive cycles after recalibration). Affirmation is just as concrete: validation requires reliability above the dual-rater κ gate, predictive lift on EBITDA realization at twelve months versus a non-VCAF control, and replication across at least five firms with four quarterly cycles each. Both standards are pre-registered (E.4, E.6, E.10).

E.1 Hypothesis Sequence

Four claims, tested in order. The order is the framework's own gating logic.

H1. VCAF can be scored reliably across raters and time. Reliability gate.

H2. VCAF measures a construct distinct from, but related to, engagement and general culture. Convergent and discriminant validity gate.

H3. Baseline EAI predicts future Adjusted EBITDA growth, retention, safety incidents, and productivity per FTE. Predictive-validity gate.

H4. ΔEAI after intervention predicts ΔOutcome on the same set. Intervention-validity gate.

Failing H1 prevents H2. Failing H2 weakens H3. Failing H3 retires the financial-translation claim per the falsification criterion in E.8. The hierarchy is not a research preference; it is the order in which a reader is entitled to demand evidence.

E.2 Study Phases

Phase 1. Instrument validation. Quarters 1 to 4 of the first deployment. Refines virtue definitions, anchors, prompts, and rubric calibration under real conditions. Dual scoring runs every cycle. Weighted Cohen's κ and Cronbach's α are computed and reported against the thresholds in E.4 before any EAI is released to operator or board. Cycles that fail thresholds trigger rubric recalibration before the next cycle, not score release.

Phase 2. Convergent validity. Begins in quarter 2 and overlaps Phase 1. Gallup Q12 administered in the same measurement window as VCAF. Denison Organizational Culture Survey norms cited rather than re-administered. Bloom and Van Reenen's World Management Survey people-management subscale referenced where firm scale and rater access permit. Convergence on shared dimensions tests that VCAF does not contradict prior validated work; per-virtue Dₖ/Pₖ resolution tests that VCAF adds diagnostic information those instruments do not capture.

Phase 3. Predictive validity. Quarters 4 to 8 of the first deployment, extending across cohort growth. Pre-registered prediction: at each firm in the cohort, EAI changes come before matching changes in Adjusted EBITDA, retention, safety incident rate, and productivity per FTE. The direction and rough magnitude of those changes match what the framework predicts. Lag structure is open at first deployment and estimated as the cohort accumulates. The linear EAI × PP form's adequacy is itself one of the questions the validation cohort answers; sigmoidal and piecewise-linear specifications (Appendix B) are tested against the cohort once n permits.

Phase 4. Intervention study. Initiated within the first deployment, formalized as the cohort grows. VCAF identifies highest-weight, highest-MI virtues. System audit per Step 6 targets the contradictions that produce the gap. Re-measurement at one and two quarters tests whether intervention closes the measured gap and whether closure is followed by outcome movement. This phase converts VCAF from a diagnostic to an intervention instrument.

E.3 Sample Design

The validation cohort: 15-30 owner-led firms across industries, quarterly measurement, four to eight quarters per firm, virtue scores paired with Adjusted EBITDA, turnover, safety incidents, productivity per FTE, and promotion ratio at each cycle.

The realistic starting design is n=1: the first deployment I run. Subsequent deployments extend the cohort. Partner operators running VCAF in their own firms extend it faster. Until the cohort reaches n ≥ 5 with at least four quarterly cycles per firm, claims are descriptive and instrumental (VCAF can be deployed and produces stable scores) rather than predictive (VCAF forecasts outcomes). The financial-translation claim is held in escrow against that threshold.

A pilot deployment inside an aligned partner firm is open as a parallel path. The first owner-operated deployment remains the primary validation commitment because the operator owns both variables (the alignment work and the EBITDA outcome) and underwrites the test with the operator's own capital. A pilot site does not match that incentive structure, so its evidence weight is lower. It is still useful for two reasons: it accelerates instrument-validation work (rubric calibration, dual-rater κ, story-prompt rotation) before the primary deployment begins, and it produces a second within-firm trajectory that the elasticity model in Appendix B requires. A qualifying pilot site is an owner-led firm whose principal commits to a four-quarter cycle, a structurally independent Dₖ rater, the pre-registration rules in E.6, and pre-registered outcome variables. Pilot data sits alongside primary-deployment data in the cohort; results are reported with deployment context flagged so a reader can weight evidence accordingly.

Smaller sample does not weaken the reliability work. Cronbach's α and weighted Cohen's κ are within-firm-within-cycle measures and are tested from quarter one of the first deployment. Predictive validity is the claim that scales with cohort size; reliability is not.

E.4 Validation Gates and Thresholds

Two statistical thresholds gate the framework. Cronbach's α measures internal-consistency reliability of a survey instrument: when multiple items are measuring the same underlying construct, employees who score high on one item should score high on the others. α = 0.80 is the convention for instruments stable enough to publish; below 0.80, the items are noisy or measuring different things. Weighted Cohen's κ measures inter-rater agreement adjusted for chance: when a human rater and an LLM rater independently assign rubric scores, κ = 0.70 means their agreement is substantially better than coincidence. Below 0.70, the rubric is too subjective or the raters are reading different signals. Both are computed cycle-by-cycle and reported alongside the EAI.

Setup. Cronbach's α runs on the Likert-survey responses for a virtue scale: each cycle, the responses across the items in that scale go into the standard α calculation, returning one number per virtue scale per cycle. The minimum sample for a stable read is roughly 30 respondents per virtue scale. Weighted Cohen's κ runs on paired human-and-AI ratings of the same stories: each cycle, every story is independently scored by both raters on each rubric dimension, and the weighted κ is computed on the pair of score sets per virtue. The minimum is roughly 20 stories per virtue with both raters scoring all of them. Standard statistical software (R, Python, SPSS) produces both numbers from the raw inputs; no specialized package is required.

Each phase has explicit pass/fail gates. Failure routes to instrument revision rather than narrative explanation.

Reliability gates. Cronbach's α ≥ 0.80 per virtue scale. Weighted Cohen's κ ≥ 0.70 between human and AI raters. Test-retest stability: scores within measurement noise across cycles at firms reporting no material culture change.

Validity gates. Convergent: VCAF correlates with Gallup Q12 on shared engagement-adjacent dimensions at conventional thresholds. Discriminant: VCAF retains variance unexplained by Q12 (Gallup's twelve-item employee engagement instrument, which measures workforce engagement at the firm level rather than alignment at virtue resolution), demonstrating that per-virtue resolution survives the comparison. Incremental: predictive models including EAI outperform models including only engagement and management-practice scores on the outcome set in E.5.

Predictive gates (cohort ≥ 5 firms with four quarterly cycles each). Within-firm ΔEAI correlates with within-firm ΔAdjusted EBITDA in direction across four cycles. Effect size meets a pre-registered threshold set before the cohort reaches the gate.

A gate failure is not a finding to explain. It triggers rubric, prompt, or specification revision before the next deployment, or a narrowing of the framework's claims if the failure persists.

E.5 Analysis Plan

Three layers, run at every reporting cycle.

Descriptive. Per-virtue Dₖ, Pₖ, VCARₖ, MIₖ. Aggregate M, σ, EAI. Concentration of misalignment by virtue. Distribution of story scores within and across virtues.

Predictive. Test whether baseline EAI and within-firm EAI changes predict later-cycle outcomes (Adjusted EBITDA, voluntary turnover, OSHA recordable incident rate, Adjusted EBITDA per FTE), with firm size, baseline margin, four-digit NAICS, and ownership-transition status held as controls.

Incremental. Compare predictive models with and without VCAF terms against models including only engagement and management-practice covariates. The test is whether VCAF adds explanatory power over instruments already in the field, not whether it correlates with outcomes in isolation.

The outcome variables are locked in before the first deployment, each with its measurement source and computation rule: Adjusted EBITDA per the build-up rules in Appendix A and B; voluntary turnover by department; OSHA recordable incident rate; Adjusted EBITDA per FTE; internal promotion ratio. Data sources are specified per firm at first deployment and locked.

E.6 Pre-Registration Commitments

The first deployment is pre-registered before story collection begins. Pre-registration locks the following decisions:

• Virtues selected and the rationale for each weight wₖ.
• Rubric anchors (Appendix C.2) and the moral-vocabulary exclusion rule.
• Story-prompt set and the rotation logic across cycles.
• Likert item bank per virtue, the rescaling formula, and the rotation cadence.
• The 60/40 narrative-Likert blend and the conditions, if any, under which it would be revisited.
• Human and AI rater identities, with model and prompt versions logged per cycle (Appendix C.6).
• λ and ρ values (Appendix B; base case λ = 0.55, ρ = 0.15) and the EAI threshold bands.
• Non-compensatory triggers: which virtues, what MIₖ threshold.
• Productivity Potential estimation method per cycle, with the default-fallback start and a committed migration to the elasticity model when the cohort supports it.
• Outcome variables, operational definitions, and data sources.
• Historical Adjusted EBITDA series for the firm: at minimum the trailing twelve months of recast figures, used as the baseline for tracking EBITDA changes across measurement cycles.
• Statistical tests for reliability, convergence, and prediction.

Pre-registration is the answer to the framework's central methodological risk: that λ, ρ, the narrative-Likert blend, the cap, the thresholds, or the outcome set could be tuned to make a given firm look good or bad. Locking those decisions before scores are computed is how the risk is neutralized. A reader is entitled to see the rules before seeing the numbers they produced.

E.7 Benchmarking Instruments and Convergent Validity

VCAF is cross-applied against established instruments to confirm the measurement converges with prior validated work on shared dimensions. Cross-applied means the same workforce, or a representative sample, completes both VCAF and the external instrument within the same measurement window, and scores on shared dimensions are compared for correlation. Convergent validity is established where VCAF and the external instrument correlate on dimensions they both claim to measure, and where VCAF adds resolution on virtue-specific alignment not captured elsewhere.

Three instruments anchor the benchmarking set.

• Gallup Q12 is administered quarterly as the convergent-validity instrument at the engagement-and-management layer. Twelve items, approximately ten minutes per employee, licensed published benchmarks. Feasible on a sub-100-person initial deployment.
• Denison Organizational Culture Survey is cited against its published norms rather than re-administered, providing a reference point for VCAF scores at the culture-trait level without the response-fatigue cost of a second full instrument.
• Bloom and Van Reenen's World Management Survey, people-management subscale, serves as a benchmarking reference for Dₖ scoring in manufacturing-context deployments. Convergence here tests whether declared management practices in VCAF track the practice-level scoring the WMS program has accumulated across more than 20,000 management interviews across 35 countries (worldmanagementsurvey.org). Cross-industry deployments use the same convergent-validity logic with industry-appropriate benchmarks.

Construct validity is supported where VCAF correlates with these instruments on shared dimensions and adds diagnostic resolution on virtue-specific alignment not captured elsewhere.

E.8 Falsification Criterion

If within-firm EAI movements do not correlate with within-firm Adjusted EBITDA changes in direction over four quarterly cycles, the linear translation fails and the framework reverts to a diagnostic-only instrument. Independent of direction-correlation, if the Pₖ measurement layer fails its reliability targets (Cronbach's α < 0.80 or weighted Cohen's κ < 0.70) across two consecutive cycles after rubric recalibration, the instrument is unfit for financial translation regardless of EAI behavior. Either failure forces a revision in scope; persistent failure across multiple deployments forces abandonment of the dollar-translation claim.

E.9 Publication Path

The framework's evidence will be staged across multiple publications: methods first, then validation and intervention findings as the cohort matures. Each publication is gated by the validation milestones in E.1 and E.4, not by a fixed schedule.

E.10 Field-Impact Criteria

VCAF justifies broader adoption only if it demonstrates three things together: it measures culture gaps at greater virtue-level resolution than existing tools; it predicts operational outcomes earlier than those tools; and it gives operators a workable lever for changing those outcomes that diagnostic-only frameworks do not provide. Any one of these in isolation produces a useful instrument. Falling short on the third (measurement and prediction without intervention validity) is the most likely outcome and would still leave VCAF a useful diagnostic. Falling short on the second (measurement without prediction) retires the financial-translation claim. Falling short on the first (measurement that does not improve on existing instruments) retires VCAF.

The dollar figures in the body of this paper carry the burden of meeting that standard.

E.11 Pre-Registration of First Deployment

The Open Science Framework (OSF) is a public registry where research designs are filed before data collection so reviewers can verify the design was not tuned to the results. The first deployment will be filed there or with an equivalent registry before story collection begins, and written up as a separate methodology paper with full rater-agreement statistics, longitudinal EAI movement, and Adjusted EBITDA correlation, intended for submission to a peer-reviewed organizational-behavior or management-science journal. Until then, VCAF is a methodology proposal awaiting validation, and the dollar figures it produces are diligence outputs to the operator deploying it, not empirical claims to the field.

E.12 Future Research: Comparative Virtue Performance Across Firms

Validated at sufficient cohort scale, VCAF would generate a dataset no existing instrument produces: per-virtue Pₖ and Dₖ scores paired with normalized EBITDA across firms, industries, and time. With enough firms and enough cycles, the data could start to answer a question the field has not been able to test directly. Do some virtues, when practiced well, drive higher operational performance than others?

This subsection sets up that future research in three parts: why the question matters, how the data would be compared, and where the assumption fails if applied too broadly.

Why the question matters. Existing culture-to-performance research (Denison; Bloom and Van Reenen; Cameron and Quinn) has tested the link between general culture quality and firm performance. None of that work resolves at virtue-level granularity, because none of the instruments scores virtues separately against declared baselines and pairs the gap with normalized profit. If a cohort of validated VCAF deployments showed, for example, that within-firm gains in Accountability EAI predicted larger Adjusted EBITDA gains than equally sized gains in Innovation EAI, the finding would change how operators allocate intervention attention. It would also give boards a defensible answer to the question "which culture investments produce the most operating leverage," a question they currently answer with consultant intuition. The same data could test whether geography (regional norms around accountability, fairness, or trust) and owner character (high-trust versus high-control operating styles) condition virtue performance the way industry does. The cohort eventually carries enough variation to test those questions if the design supports them.

How the comparison would be structured. Three layers, in increasing strength of claim.

The first layer is within-firm, longitudinal: across a single firm's measurement history, do gains on virtue A's EAI precede larger Adjusted EBITDA gains than equally sized gains on virtue B's EAI? This layer is internal-to-the-firm and controls for industry, scale, and ownership context. It produces firm-specific findings about which virtues carry the most operating weight at that firm, which is itself a useful intervention-design output.

The second layer is within-industry, cross-firm: holding industry constant, does the virtue with the largest within-firm coefficient cluster across firms? If specialty B2B manufacturers consistently show that within-firm ΔAccountability_EAI drives within-firm ΔAdjusted EBITDA more than within-firm ΔFairness_EAI, that pattern is a candidate industry finding. The instrument has to clear convergent and discriminant validity gates (E.4) before this layer is interpreted.

The third layer is cross-industry: do the virtue rankings replicate across industries, or do they reorder? This is the layer where most of the danger lives, treated below.

Why the assumption fails if applied too broadly. Industries are not interchangeable measurement environments. A specialty manufacturer's operating performance is most exposed to safety practice; a regulatory failure on the floor produces direct workers'-comp, OSHA, and downtime cost. An accounting firm's operating performance is most exposed to integrity practice; a single audit lapse can produce a malpractice claim that exceeds annual revenue. A SaaS company's operating performance is most exposed to commitment-keeping with customers and engineers; a missed reliability target compounds across renewal cycles. Reading "virtue X drives performance more than virtue Y" as an industry-agnostic claim would be a category error.

The right framing is that environments select for virtues differently, and the same virtue can carry different operating leverage depending on what the work actually is. The cross-industry comparison's job is to surface where the environment-conditioning is, not to declare a universal ranking. A cross-industry finding that Accountability outperforms Fairness at some statistical threshold is not a recommendation that every firm prioritize Accountability; it is a hypothesis that the operating channels for these two virtues differ in size and that the differential is detectable at cohort scale. The actionable claim still has to be made firm by firm, with the firm's industry and operating model in the frame.

Methodological guardrails. Three rules govern any virtue-performance comparison VCAF data feeds.

• Industry-conditional reporting. All cross-firm virtue-performance comparisons are reported with industry held as a covariate, never aggregated into industry-agnostic rankings. The headline output is the industry-by-virtue interaction, not the virtue main effect.
• Mechanism-specified prediction. For any virtue claimed to drive operating performance in a specific industry, the operating channel has to be named in advance, before the data is read. Safety in manufacturing runs through workers'-comp, EMR, and downtime; integrity in accounting runs through malpractice and re-engagement loss; reliability in software runs through customer churn and renewal compounding. The same logic anchors interventions: an intervention without a named channel is hope, not a tested mechanism. A finding without a named channel is descriptive, not causal.
• Adversarial replication. Hypotheses are pre-registered before the cohort is read. Null results are reported with the same prominence as positive results. The framework's credibility depends on its willingness to publish failures of its own predictions.

What this enables. If the work clears those gates, VCAF data would let the field move past general culture-to-performance correlations and toward virtue-by-environment specificity. That is a research arc, not a deliverable; the paper makes no claim that VCAF currently supports it. The first pre-registered deployment, plus subsequent cohort firms, accumulate the evidence base. Until then, this subsection sets the question and the rules under which it could be answered.

---

Appendix F: Deployment Runbook

Phase 1 runs ninety days because Watkins (2003) frames that period as the window where a new operator can listen long enough to hear the system without losing the authority to act on what they hear. Both directions of error are costly: a shorter window leaves the diagnosis thin, and a longer one lets the contradictions the operator should have caught harden into facts the operator now owns.

F.0 Operational Prerequisites

VCAF assumes three operational resources are in place before Day 1. Each prerequisite is named alongside the F.1 move that depends on it, so an operator can see what breaks if a resource is missing.

• Dₖ reader. Anyone outside the assessed leader's reporting chain: a board member, an outside advisor, or a rotating internal rater drawn from a peer function. F.1 Move 0 triangulation requires this reader.
• Anonymous story intake. Anonymous on the submitter side; an off-the-shelf survey tool at small-firm scale satisfies it, and a vendor relationship is not required. F.1 Move 1 (locate the contradiction) depends on the candor that the intake's anonymity protects.
• Frontier-model LLM rater. Accessed through the operator's standard tool stack: a consumer chat interface, a plugin, or an API. The load-bearing requirement is logging the model identifier and prompt version per cycle, since F.1 Move 6 (re-measure and ratify) depends on rater reproducibility across cycles. The framework does not assume custom software.

Substituting peer-rater pairs for the Dₖ reader or in-person interviews for the intake channel breaks the reliability gates in Appendix E.4 and should be documented as a deviation rather than treated as equivalent.

The intervention plan VCAF produces is independent of the financial translation only after Phase 1 reliability gates clear. Until Cronbach's α ≥ 0.80 and weighted Cohen's κ ≥ 0.70 are met across two consecutive cycles, intervention design is treated as exploratory; the routing claim from Step 6 is held conditional on those gates. Where reliability has not yet cleared, the intervention layer is reported as candidate-routing pending validation.

Where the manager being assessed is the source of misalignment, the anonymity question becomes structural. Third-party administration, theme-level reporting, and private-by-submitter marking are operationally required at small-firm scale to keep the signal honest; an operator who waives any of these discloses the change to the workforce before the next collection cycle so employees can write or abstain accordingly.

The framework's full quantitative form assumes organizational depth; reliability targets in Appendix E.4 are calibrated against firms with thirty or more story submissions per virtue per cycle. Smaller teams may run a narrative-only adaptation of the same architecture, with quantitative targets adjusted accordingly; full quantitative deployment below that threshold is reported as directional pending broader cohort calibration.

Phase 1, Days 1 to 90. Listening.

Weeks 1 to 2. Care first, then introductions. The workforce has just had its daily life rearranged by a change they did not initiate. Asking them in week one to submit narrative accounts of how the prior owner handled safety, fairness, accountability, and other virtues is intrusive if the frame is wrong. Weeks 1 to 2 are therefore shift-by-shift introductions, named commitments about what will not change in the first 90 days (pay, roles, benefits, reporting lines), and a plain-language description of the story-collection process, its purpose, who sees what, and the right to decline. No story collection begins until week three.

Consent is explicit. Participation in story collection is voluntary and declining carries no operational consequence. In a small workforce, full anonymity in narrative collection is operationally impossible: everyone can infer who said what. The design acknowledges this directly with the workforce rather than promising an anonymity the structure cannot deliver, leans harder on observable behavioral data (turnover, safety incidents, exit interviews, Glassdoor history) as the primary Pₖ input, and treats narrative as supplementary at this scale. Employees who opt out contribute to Pₖ through the behavioral channel only; no score penalty attaches to non-participation.

Owner access to stories. Default: the incoming owner sees rater-coded scores and thematic summaries, not raw submissions. Raw text is held by the outside rater team and is used for scoring, calibration, and theme extraction. If an owner decides that direct reading of raw stories is operationally necessary, that choice is disclosed to the workforce before the next collection cycle so employees can write or abstain accordingly. The default protects the candor the instrument depends on; the disclosure rule keeps the operator's authority intact without misrepresenting the arrangement to the people submitting stories.

Private-by-submitter marking. Employees may mark a story private at submission. Private stories are read and scored by the rater team, count toward the virtue's Pₖ at the dimension level, and are excluded from thematic summaries returned to the owner. Rater-only visibility lets the signal still count without exposing the individual narrative. If private-marked submissions exceed 20 percent of a virtue's stories in a cycle, the rate itself is logged as a Pₖ-lowering signal on that virtue and surfaced in the theme report as an environmental finding about candor on that dimension.

Weeks 3 to 6: Story collection and Likert survey administered through a third party. Concurrent observational data (trailing 12-month turnover, safety incident log, promotion history). The Adjusted EBITDA baseline is established during Phase 1 financial review; Weeks 3 to 6 verify the baseline against current books and flag any divergence between the prior baseline and the operator's own read.

Weeks 7 to 10: Story scoring (human + AI) and survey analysis. Behavioral triangulation.

Weeks 11 to 13: Draft Pₖ scores per virtue. Confirm Adjusted EBITDA baseline. Do not declare Dₖ yet.

Phase 2, Days 90 to 180. Baseline and commit.

Weeks 14 to 16: Declared values framework finalized with outside rater scoring Dₖ. For mid-market owner-operated firms, this governance can be minimal: owner plus one trusted operator plus one outside advisor. Function matters more than formality. Full VCAF run on Adjusted EBITDA.

Weeks 17 to 20: Findings shared with the team as a commitment to act, not as a performance review of the workforce.

Weeks 21 to 26: System audit on the priority virtues, ordered by weighted MIₖ. Bandwidth caps the count: two to four in a quarter is typical, more if the contradictions share a common system (one comp plan, one cadence) and one structural change closes several at once. Contradictions identified and remediation scoped.

F.1 Intervention Mechanics

Step 6 of the body declares the principle (eliminate contradictions between stated values and actual systems). This subsection walks an operator from a finished VCAF run to a closed gap on a single virtue. The sequence is seven moves: triangulate the signal, find the contradiction, test it against the evidence, choose the lever by leverage rather than visibility, set predictions before the change is announced, announce the change with the contradiction named, then re-measure and either ratify or revise. The first move (triangulate) is a prelude that protects the rest from acting on noise; the next six are the operator's intervention arc. The example uses Arete's Accountability finding (Pₖ 54, Dₖ 80, MIₖ 0.26, top-priority virtue) so the reader can trace it back to the worked numbers in Appendix D.

Across every move, three framing constraints stay in view. The lever hierarchy: structural changes (comp plans, promotion criteria, review cadence) outperform procedural changes (new meeting templates, RACI charts), which outperform communication-only changes (values training, all-hands speeches); Move 3's mechanism choice is read against this hierarchy. The treatment unit: comp plans and promotion criteria sit at firm-level, review cadence at department- or team-level, supervisor coaching at individual; Move 1 carries a treatment-unit note so the next cycle's measurement is read at the right resolution. The time-to-signal: Likert pulse moves in one to two quarters, narrative themes take two to three, operating metrics range from one quarter (cycle-time, turnover) to multiple years (EMR, regrettable-attrition trend). The Move 4 prediction window matches the slowest of the three predicted layers.

F.1 produces operator hypotheses tested against the next measurement cycle. Move 6 either ratifies the hypothesis or routes the operator back to Move 1; intervention failure inside the framework is a Move 1 finding.

Move 0. Triangulate before acting. Before walking the candidate mechanisms in Move 1, confirm the misalignment signal across at least two independent data sources. Likert items, story narratives, and observable operating data (turnover, safety incidents, promotion ratios) are the most-available sources at small-firm scale. A virtue's MIₖ that surfaces in one source but not the others points to a measurement issue rather than a routing target; the rubric or item set is recalibrated before intervention design begins. Triangulation keeps the framework from acting on noise.

Move 1. Locate the contradiction. A low Pₖ on a virtue the firm declares is almost never a people problem. The people are operating inside a system that rewards, punishes, signals, and selects on different things than the declared virtue. Walk those systems first. For Accountability at Arete the candidates are review cadence, promotion criteria, the comp plan, the meeting structure, and the public response when a forecast misses. Rank the candidate mechanisms (review cadence, comp plan, promotion criteria, meeting structure, public response when a forecast misses) by which one is doing the most work against the declared virtue. If review cadence is monthly and forecast misses are discussed only when the miss is large, the cadence punishes raising small risks early.

The contradiction usually classifies as one of five fault modes: incentive misalignment (the comp plan or promotion criteria reward what the declared virtue contradicts), information breakdown (the system that should surface the signal does not), leadership inconsistency (the modeled behavior contradicts the declared value), process contradiction (the operating cadence punishes the behavior the declared value requires), or workforce disengagement (the trust contract is broken and the workforce no longer responds to system signals; structural changes alone do not move Pₖ until the trust work happens first). Naming the fault mode at Move 1 narrows the Move 3 lever choice; the disengagement case carries an explicit sequencing rule, with trust-rebuilding work preceding structural change.

Move 2. Test the contradiction against the evidence. Cross the candidate against the qualitative and quantitative inputs. The Arete Accountability stories show that public mistake-owning happens when someone owns a defect, but the Likert items on raising forecast risks without penalty and on review-cadence signal score lowest. The two evidence streams point at the same mechanism: people will own a clear failure after the fact, and they will not raise a soft risk before it. The contradiction is now locatable to the review cadence; Move 3 chooses the lever to fix it.

Move 3. Choose the intervention by leverage, not by visibility. Apply the lever hierarchy from the framing notes above. For Arete, the structural move is to change review cadence to weekly and to add an explicit forecast-risk surfacing item to the agenda that produces no penalty for raising. The procedural move is the new agenda item. The communication move is naming the change in the Weeks 17 to 20 commitment session as a response to the score, not as a generic improvement initiative. Watch for second-order effects when the chosen lever interacts with another virtue's standing: accountability tightening without psychological safety produces fear rather than ownership; raising near-miss reporting cadence without a no-penalty norm collapses reporting volume. Move 1's mechanism choice is read against the firm's standing on the supporting virtues.

Move 4. Pre-commit to a measurable outcome at three levels. Before the change is announced, the operator names what the next cycle's data should show across three independent layers. The Likert prediction names which items move and by how much on the 1-to-5 scale. The operating-metric prediction names which line item from the F.5 anchor set moves and in what direction. The story-signal prediction names which narrative pattern should appear or disappear in the next collection. For Accountability at Arete, the predictions are that the two lowest-scoring Likert items move at least half a point within two cycles; that voluntary IC turnover trends down or that forecast accuracy on the rolling four quarters trends up; and that story narratives begin to include raised-risk events with no negative consequence. All three predictions met is success. Partial success at two-of-three triggers a Move 1 review of which prediction layer failed and why. One-of-three or worse counts as intervention failure and escalates to Move 6.

Move 5. Announce the change with the contradiction named. Findings are shared as a commitment per Weeks 17 to 20, and the contradiction is stated plainly. The script is structurally simple: "We declare accountability. We were running a review cadence that punishes raising risks early. We are changing the cadence and the agenda. The data we expect to see next cycle is X. We will report on whether we saw it." Naming the contradiction is what separates an intervention that changes the system from one that updates the values poster.

Move 6. Re-measure and ratify or revise. At the next quarterly run, collect a fresh round of stories on the targeted virtue, refresh the Likert items per the rotation, and review whether the F.5 operating metric the operator pre-committed to has moved. Re-measurement requires a new collection cycle. Cached inputs do not satisfy Move 6. If the predictions in Move 4 were met across all three layers, the system change is ratified, and the next intervention targets the next-highest MI. If the predictions were missed, the operator returns to Move 1 with the new evidence: the contradiction was probably elsewhere, or the chosen lever was procedural when structural was needed. A failed intervention is a finding, not a defeat; persistent failure across two consecutive cycles is escalated to a system-level review.

The default sequence for two-virtue interventions in Weeks 21 to 26 is to run Move 0 through Move 5 on both virtues in parallel and re-measure at Quarter 3. Sequential intervention is an alternative when the two contradictions share a system (a single comp plan that punishes both Accountability and Fairness), in which case one structural change is tested before a second is layered.

As the validation cohort accumulates, intervention success rate (three-of-three Move 4 predictions met at Move 6, divided by total interventions attempted) becomes a published cohort-level metric, reported alongside reliability and predictive-validity statistics in Appendix E.

F.2 Cadence

The framework's discipline depends on its rhythm. Run too rarely and intervention drifts; run too often and signal collapses into noise.

Year 1 of deployment. Phase 1 listening produces draft Pₖ at Week 13. Phase 2 produces the first full VCAF run at Week 16. Phase 2 cycle 1 produces a re-measurement at the end of Quarter 3. The first quarterly cycle in production form runs at the end of Quarter 4. The first year therefore carries one full instrument build (Dₖ and Pₖ scored together for the first time) and three measurement points. The first measurement is the build itself; the next two test what the build claimed.

Year 2 onward. Full VCAF runs (Dₖ scored by the outside rater, Pₖ scored by stories plus Likert) move to semi-annual cadence. Pₖ-only pulse runs (one channel only, either stories or Likert, at half the volume of a full run) run at the off-quarters to detect drift between full runs. Pulses signal whether a full run is warranted before the next scheduled cycle; they do not substitute for one. Total cadence: two full instruments and two pulse measurements per year. This is the load a small workforce can carry without survey fatigue degrading response quality.

Event-driven re-measurement. A full VCAF run is rare and costly: collection burden on the workforce, calendar burden on the operator, analytical burden on the rater. The default cadence is the calendar above. An off-calendar run triggers only when something has happened that the calendar would miss in time: a leadership change at a senior level, a comp-plan rebuild, M&A activity that materially changes headcount or reporting, a recordable safety incident, a public regulatory citation, or the loss or gain of a customer that exceeds 15 percent of revenue. Off-calendar runs catch the moments when the firm has changed underneath that cadence.

Reporting alignment. Quarterly summaries land in board packets next to the financials. Full instrument outputs land in the semi-annual or annual board review. The dashboard separates the diagnostic (per-virtue scores, MIₖ, σ, EAI) from the financial translation (EBITDA at Risk, EV at Risk if at diligence or exit) so the board reads the operating signal first.

Why this cadence. Cadence is bounded by what the data, the rotation pools, and the workforce can sustain. The data needs time: a cadence-and-comp change at Move 3 takes one to two quarters to register in Pₖ, so running monthly would measure noise rather than effect. The rotation pools have limits: stories and Likert items rotate so employees do not see identical instruments cycle after cycle, and running too often exhausts the supply. The workforce has limits too: a small workforce cannot sustain four full instruments a year without people writing shorter, less candid stories. Quarterly pulse plus semi-annual full runs keeps the signal honest and the workforce willing to produce it.

F.3 Team-Level Deployment

VCAF is described in the body as a firm-level instrument because the underwriting use case is firm-level. The same architecture supports segmentation below the firm, which becomes useful as the workforce grows past the point where the owner can read the floor by walking it.

What can be segmented. The unit of analysis can shift from firm to department, shift, physical site, or function. The math does not change: per-virtue Dₖ and Pₖ are computed inside the segment; segment-level VCARₖ, MIₖ, and EAI follow from the same equations in Appendix B. The financial translation does not segment cleanly (Adjusted EBITDA is rarely allocable to a single team without contested cost-allocation choices), so segment-level outputs are reported as alignment metrics, not dollar figures.

Two comparisons. Segment-versus-firm compares one team's EAI to the firm's EAI and surfaces local pockets of misalignment that firm-level scoring would average away. Segment-versus-segment compares teams against each other and surfaces structural differences (a shop floor running tighter on Safety than the back office running loose on Accountability) that argue for differentiated intervention. Both comparisons are diagnostic, not evaluative; segments are not in competition with each other.

Use cases. Targeted intervention: when firm-level Accountability is misaligned, a segment-level read shows whether the gap concentrates in one department or distributes across the firm, which determines whether the Move 3 lever is local (a department-specific cadence change) or global (a firmwide comp-plan revision). Promotion of latent strength: a segment with raw VCARₖ > 1.0 on a virtue is a candidate for codifying its practices into the firm-level doctrine (see the worked walk-out in Appendix B). Cross-functional friction: a wide gap between Operations and Sales on the same virtue often signals an interface problem (handoff, shared metrics) rather than a culture problem in either function.

Caveats specific to segmentation. The reliability targets in Appendix E.4 are calibrated for firms reporting at least thirty story submissions per virtue per cycle. A segment that falls below that threshold cannot anchor reliability the same way the full firm can: per-segment Cronbach's α and weighted Cohen's κ are likely to fail their thresholds, and segment scores in that range should be reported as directional, with the routing claim held conditional on broader cohort accumulation. Identification risk rises with smaller units: stories collected at the team level are often identifiable to the team's manager, and sometimes to the individual. Private-by-submitter marking (Weeks 1 to 2 block above) is operationally required at this scale, not optional. Public segment ranking is not advised; the framework is a diagnostic for the operator, not a leaderboard for the workforce. Reporting protocol: segment scores are shared with segment leadership for intervention design, and only the firm-level rollup is shared with the workforce.

When to segment. Segmentation begins in Year 2, once the firm-level baseline exists. From Year 2 forward, the segment view operates as a comparison instrument against the firm baseline; standalone segment-level diagnoses still wait on cohort-level segmentation work.

F.4 Operating Cadence Beyond Year 2

After eight measurement points across two years, operating cadence is set by the data rather than by the calendar. Firms that have hit EAI > 0.85 with stable σ across four consecutive cycles move to annual full instruments and quarterly pulses. Firms still working through structural interventions retain the Year 1 cadence until two consecutive cycles show movement on the targeted virtue. The default is to slow the cadence as the firm earns it, not to anchor on quarterly forever.

F.5 Virtue-to-Mechanism Crosswalk

F.1 walks the operator from finished score to closed gap. F.5 is the starting candidate set behind Move 1 (Locate the contradiction): a library of operating mechanisms where misalignment most often surfaces, paired with the operating-metric anchors that move when those mechanisms are fixed and the time-to-signal a cycle should expect. The library proposes; the stories localize. The Likert × story routing in Step 6 narrows the candidate list to the two or three mechanisms the firm's narrative points at; F.5 is the menu that makes that narrowing tractable.

Each virtue carries a function-level read line: the one-sentence read of what the surveys-and-stories layer is reporting at the function-and-mechanism resolution. The function-level read is what makes the routing claim concrete; without it, the per-virtue mechanics read like a generic diagnostic with new vocabulary.

The crosswalk turns EBITDA at Risk from a single number into a directed claim: a virtue's MIₖ points to specific systems to walk, those systems point to specific operating-metric line items, and the operator pre-commits to which of those line items the next measurement cycle should move.

The three Arete virtues (safety, accountability, fairness) are populated against the worked-example inputs from Appendix D. Four field-typical virtues from Sull, Sull, and Chamberlain (2020), the most-cited values across 562 large U.S. firms, are populated in starter form: enough mechanism and metric content to begin a Move 1 walk in a non-Arete context, with the expectation that the operator will refine the entries against the firm's narrative during Phase 1. The architecture extends to other industries with industry-specific benchmarks. The mechanisms in the entries below are drawn from the Arete worked example, where the firm is a small specialty manufacturer; in another industry, the mechanism list shifts (a service firm watches scheduling and supervisor coaching where a manufacturer watches LOTO discipline and equipment-maintenance cadence) but the routing logic is identical.

Safety (Arete weight 0.45; non-compensatory)

Mechanisms. Pre-shift safety briefs; near-miss reporting (submission volume and manager response time); supervisor coaching cadence on safety behaviors; equipment lockout-tagout discipline; contractor onboarding; root-cause analysis on every recordable incident.

Operating-metric anchors. Experience modification rate (EMR); Total Recordable Incident Rate (TRIR, the OSHA-published recordable rate); lost-time incident rate; near-miss-to-recordable ratio; workers' comp premium per FTE; downtime hours from safety events.

Function-level read. What the surveys-and-stories layer is reporting at function resolution: which shift, which line, which supervisor cadence. Safety stories localize to specific shifts and procedures; the routing identifies whether the Move 3 lever is shift-level (supervisor coaching) or firm-level (equipment LOTO discipline).

Notes. Non-compensatory virtue. Intervention triggers regardless of composite EAI when MIₖ > 0.30. After the fix, lost-time rate moves within two quarters; EMR carries a three-year tail (the policy-year carry from Step 5 is the lag).

Accountability (Arete weight 0.30; highest weighted MIₖ in the Appendix D worked example, hence the priority intervention target after the Safety gate clears)

Mechanisms. Review cadence (weekly versus monthly); forecast-risk surfacing as a standing agenda item that produces no penalty for raising; comp-plan tie to commitments kept versus commitments shifted; promotion criteria (results-as-promised versus visibility); explicit escalation paths; root-cause discipline when results miss.

Operating-metric anchors. Voluntary turnover among individual contributors; forecast accuracy on the rolling four-quarter view; project cycle time at handoff stages; customer rework and warranty claim rate; defect root-cause closure count per quarter.

Function-level read. What the surveys-and-stories layer is reporting at function resolution: which decision pattern (forecasting, escalation, post-mortems) the firm fails on. Accountability stories localize to specific decision moments; the routing identifies whether the Move 3 lever is review-cadence (department-level) or comp-plan (firm-level).

Notes. The F.1 worked example uses Accountability. The candidate mechanism walked at Move 1 is review cadence, on the argument that monthly review punishes raising small risks early. The Move 4 pre-commitment is that the two lowest-scoring Likert items move 0.5 on the 1-to-5 scale within two cycles. After the fix, Likert pulse moves in one to two quarters; narrative takes two to three.

Fairness (Arete weight 0.25)

Mechanisms. Compensation-banding transparency where present, or salary-decision documentation where banding is absent; promotion-decision documentation; scheduling discretion (overtime distribution, shift assignment); discipline consistency across roles and shifts; complaint channel and follow-through; performance-review calibration across managers.

Operating-metric anchors. Regrettable-attrition rate (high performers leaving); engagement-survey items concentrated on promotion and pay; internal-promotion ratio; EEOC charge rate; Glassdoor or industry-board rating drift.

Function-level read. What the surveys-and-stories layer is reporting at function resolution: which decision class (promotion, scheduling, compensation, discipline) the firm fails on. Fairness stories typically concentrate in one decision class rather than spreading; the routing identifies the class first.

Notes. Fairness gaps often concentrate in one mechanism (typically promotion decisions) rather than spread across the list. Smaller firms usually lack formal banding; at this scale the fix is documentation discipline, not banding redesign. After the fix, the next signal is the next promotion or compensation decision, two to three quarters out.

Integrity (starter; Sull rank 1 at 65%)

Mechanisms. Documented commitments to customers, suppliers, and employees; consistency between public statements and internal practice; an ethics-channel concern process; visible consequences when leadership makes a call that contradicts a stated value.

Operating-metric anchors. Customer churn flagged by trust events; supplier on-time-pay rate; ethics-channel report rate and time-to-resolution; PR or regulatory event count.

Function-level read. What the surveys-and-stories layer is reporting at function resolution: which trust contract the firm is breaking (customer, supplier, employee, regulator).

Notes. Starter. Field-typical virtue from Sull, Sull, and Chamberlain (2020) top-of-distribution. Integrity is the most generic value in the field and the most likely to read as boilerplate without behavioral anchors at Step 1; the Pₖ stream is doing the discriminating work here.

Collaboration (starter; Sull rank 2 at 53%; one of Sull's four negative-correlation values)

Mechanisms. Cross-functional handoff rituals; shared objectives between functions; cross-team review cadence; recognition for cross-function contribution; explicit escalation paths between functions.

Operating-metric anchors. Project lead-time at handoff stages; cross-team Likert items in the survey rotation; internal-NPS between functions; on-time interface delivery rate.

Function-level read. What the surveys-and-stories layer is reporting at function resolution: which interface (Sales-Operations, Engineering-Product, Service-Sales) the friction concentrates on.

Notes. Starter. One of the four values Sull et al. (2020) found negatively correlated with employee experience: firms that publish "collaboration" tend to score worse on it than firms that do not. The Pₖ stream is load-bearing on declared-collaboration calls.

Customer Focus (starter; Sull rank 3 at 48%; one of Sull's four negative-correlation values)

Mechanisms. Voice-of-customer cadence in operating reviews; NPS or CSAT pulse on a cycle the firm can act on; customer-escalation routing and SLA; post-sale follow-up discipline; sales-engineering coordination on commitments made versus deliverable.

Operating-metric anchors. Net Promoter Score; Customer Satisfaction; gross retention; expansion revenue; complaint-to-resolution time.

Function-level read. What the surveys-and-stories layer is reporting at function resolution: whether the practice gap concentrates in pre-sale (need-understanding) or post-sale (commitment-keeping).

Notes. Starter. Sull et al. (2020) negative-correlation flag applies. Customer-focus declarations correlate negatively with employee-reported customer focus, often because the practiced behavior is order-taking rather than need-understanding (the same drift the operator anecdote in Values Versus Virtues describes).

Respect (starter; Sull rank 4 at 35%)

Mechanisms. Manager 1:1 cadence; response time on internal requests; tone in escalation; public-recognition norms; meeting attendance and cancellation patterns.

Operating-metric anchors. Pulse engagement on respect items; voluntary-quit clustering by manager; escalation volume by team; internal-NPS between manager levels.

Function-level read. What the surveys-and-stories layer is reporting at function resolution: which manager (or which manager class) the gap concentrates around.

Notes. Starter. Field-typical virtue from Sull, Sull, and Chamberlain (2020). At small-firm scale the per-manager segmentation in F.3 is usually the right read because variance across the firm is small but variance across managers can be large.

At Move 1 of F.1 the operator walks the candidate mechanisms above for the priority virtue (highest weighted MIₖ). At Move 4 the operator pre-commits to the three-level prediction (Likert, operating-metric, story signal) using these anchors. F.5 is the bridge from a per-virtue MIₖ to the systems to walk, the metrics to watch, and the function-level read of what the firm's narratives are reporting.

---

Appendix G: Literature Positioning and References

VCAF inherits from four research traditions that the section names by source: organizational-learning theory and institutional decoupling, climate and culture measurement, coordination-cost economics, and virtue ethics. Each tradition supplies a piece of the framework's load-bearing theoretical scaffolding. Subsections G.1 through G.4 walk the lineages; subsection G.5 lists the references.

G.1 Espoused Theory and Decoupling Lineage

The declared-versus-practiced distinction VCAF operationalizes through Dₖ and Pₖ has its strongest theoretical antecedent in Argyris and Schön's (1974) introduction of espoused theory and theory-in-use as paired constructs. Espoused theory is what professionals say guides their action; theory-in-use is what does. Argyris and Schön (1978) extended the distinction from individual professional practice to organizational learning, treating the gap between the two as the diagnostic locus for revising the assumptions an organization runs on at the level of practice. VCAF carries that frame to virtue resolution, with a baseline-aware metric structure: the VCARₖ ratio captures how much declared intention is realized; MIₖ captures the absolute gap; the two readings together preserve information either alone would lose. The intervention layer also differs from Argyris and Schön's Model II reflective practice and double-loop learning. VCAF routes through the contradiction map in Step 6, identifying which management system is producing the gap and changing it structurally.

The institutional-theory analog of espoused-versus-enacted is decoupling, which Meyer and Rowan (1977) introduced to describe organizations whose formal structures arise as reflections of rationalized institutional rules rather than from technical demands. In their account, formal structures and ongoing activities are decoupled, and "in place of coordination, inspection, and evaluation, a logic of confidence and good faith is employed." Bromley and Powell (2012) extended the framework, arguing that the spread of rationalized standards across modern organizations creates pervasive decoupling between formal claims and enacted behavior in domains from environmental compliance to employee well-being. VCAF operationalizes a firm-level analog of the same gap. The institutional literature reads decoupling primarily at the firm-environment boundary and treats it as ceremonial; VCAF reads it inside the firm at virtue resolution and measures it through paired Dₖ and Pₖ scores aggregated through EAI, with the gap translated into operating-profit terms via the back-solve and the per-virtue mechanism walk-out (Appendix D Tables D.5a and D.5b). The contribution claim is empirical: VCAF surfaces decoupling at the resolution and in the units where an operator can act on it.

G.2 Climate, Culture, and Practice Instruments

Several established instruments anchor VCAF's positioning at the climate, culture, and management-practice layer.

Kaptein's Corporate Ethical Virtues model (2008) is the closest antecedent, introducing a multi-virtue diagnostic that has been validated across dozens of subsequent studies; VCAF's contribution relative to CEV is the explicit declared-versus-practiced split at virtue resolution and the coupling to EBITDA-based financial output. Cameron and Quinn's Competing Values Framework (2006) is a parallel weighted cultural diagnostic widely used in organizational research; CVF is purely diagnostic, and VCAF differentiates the same way it does against CEV, by adding a declared-versus-practiced split at virtue resolution and a financial output layer. Victor and Cullen's (1988) ethical climate work grounded a research tradition showing that perceived ethical environment predicts employee behavior and outcomes; VCAF inherits the perception-as-data premise and extends it to declared-versus-practiced measurement at virtue resolution.

Denison's (1990) culture-to-performance research linked cultural traits to firm-level financial performance and established the empirical case that culture variables predict business outcomes. Sull, Sull, and Chamberlain (2020) studied 562 large U.S. firms with official values statements and found no correlation between published values and how well employees said the firm lived them; on four of nine tracked values (collaboration, customer orientation, execution, diversity), the correlation was negative. The Sull finding is the empirical anchor for VCAF's negative-correlation flags on the field-typical virtue starters in F.5 and one of the motivations for the framework's commitment to measuring practiced behavior alongside declared values.

Bloom and Van Reenen's World Management Survey provides the closest methodological precedent: structured, rubric-scored assessment of firm-level practices, independently rated, with inter-rater reliability as a published validation target. Bloom and Van Reenen's (2007) original 732-firm survey across four countries established the design; the World Management Survey program has since extended that methodology to more than 20,000 management interviews across 35 countries (worldmanagementsurvey.org), with identification-credible evidence, including randomized trials (Bloom et al. 2013), that differences in management practice translate to measurable differences in productivity and profitability. VCAF extends that architecture from operational management to the virtue layer beneath it, at resolution and deployment scale the WMS instrument does not target. Schein's (2010) three-level culture model (artifacts, espoused beliefs and values, basic underlying assumptions) provides the theoretical grounding for the Dₖ/Pₖ split; the framework's measurement layer reads at Schein's espoused-beliefs level for Dₖ and at the underlying-assumptions level for Pₖ.

McKinsey's Organizational Health Index (Keller & Price 2011) operates at a comparable financial-linkage ambition with far larger benchmark data. VCAF differs at virtue resolution OHI does not target, in baseline-aware metrics that distinguish between firms whose absolute gaps look identical but whose declared baselines do not, and at a deployment footprint a small to mid-sized firm can carry without OHI-scale infrastructure. The owner-operator transition window is one canonical application; the architecture extends to any firm where the alignment-to-EBITDA channel is operative, with the boundary conditions named in Appendix B. OHI's published outcomes use EBITDA/sales as the financial-performance measure (top-quartile organizations 68% above-median, bottom-quartile 31%; McKinsey OHI Public Sector Practice brief), grounding VCAF's EBITDA framing in the closest comparable benchmark.

Gallup's Q12 is the dominant employee-engagement instrument in the field. VCAF positions Q12 as a convergent-validity reference at the engagement layer (Appendix E.7) while differentiating on the virtue resolution Q12 does not target.

Watkins's (2003) framework for the first 90 days after a leadership transition informs the listening-phase structure in Appendix F. The Phase 1 ninety-day window in F.0 derives from Watkins's transition research on the trade-off between listening duration and the operator's authority to act on what listening surfaces.

G.3 Coordination-Cost Foundations

VCAF's coordination-cost claim sits at the intersection of three research traditions. Transaction-cost economics (Williamson 1975, 1985; Becker & Murphy 1992) treats coordination cost as the constitutive feature of firms and locates the productivity bridge between specialization depth and coordination capacity. Psychological-safety research (Edmondson 1999, 2019) supplies the empirical mechanism by which information flow within firms is suppressed under misalignment and bad news travels late or not at all. Trust research (Zaheer, McEvily & Perrone 1998; McEvily, Perrone & Zaheer 2003) treats trust as an organizing principle that scales from interfirm to intrafirm coordination and connects directly to VCAF's Dₖ/Pₖ split. The microfoundational synthesis in Appendix B (Theoretical Foundations: Coordination Friction and Operating Profit) develops each tradition's contribution to VCAF's alignment-to-EBITDA channel.

G.4 Virtue Ethics Lineage

The virtue-ethics tradition supplies the philosophical scaffolding for VCAF's choice of virtue as the unit of measurement. Aristotle's Nicomachean Ethics establishes virtue as habituated capacity acquired through repeated action and shaped by the management systems that reward and select on it. MacIntyre (2007 [1981]) recovers the Aristotelian frame for the modern setting and supplies the practices, internal goods, and unity-of-a-life concepts that connect virtue ethics to organizational behavior. The philosophical foundations are developed in Appendix H (Philosophical Foundations: Virtue Ethics from Aristotle to MacIntyre).

G.5 References

1. Argyris, C. & Schön, D. A. (1974). Theory in Practice: Increasing Professional Effectiveness. San Francisco: Jossey-Bass.
2. Argyris, C. & Schön, D. A. (1978). Organizational Learning: A Theory of Action Perspective. Reading, MA: Addison-Wesley.
3. Aristotle. Nicomachean Ethics, Book II. (trans. W. D. Ross).
4. Becker, G. S. & Murphy, K. M. (1992). The division of labor, coordination costs, and knowledge. Quarterly Journal of Economics, 107(4), 1137 to 1160.
5. Bloom, N. & Van Reenen, J. (2007). Measuring and explaining management practices across firms and countries. Quarterly Journal of Economics, 122(4), 1351 to 1408. (World Management Survey.)
6. Bloom, N., Eifert, B., Mahajan, A., McKenzie, D., & Roberts, J. (2013). Does management matter? Evidence from India. Quarterly Journal of Economics, 128(1), 1 to 51.
7. Bromley, P. & Powell, W. W. (2012). From smoke and mirrors to walking the talk: Decoupling in the contemporary world. Academy of Management Annals, 6(1), 483 to 530.
8. Burnet, J. (1930). Early Greek Philosophy (4th ed.). A. & C. Black.
9. Cameron, K. S. & Quinn, R. E. (2006). Diagnosing and Changing Organizational Culture: Based on the Competing Values Framework (Rev. ed.). Jossey-Bass.
10. Denison, D. R. (1990). Corporate Culture and Organizational Effectiveness. Wiley.
11. Denison, D. R. & Mishra, A. K. (1995). Toward a theory of organizational culture and effectiveness. Organization Science, 6(2), 204 to 223.
12. Edmondson, A. C. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2), 350 to 383.
13. Edmondson, A. C. (2019). The Fearless Organization: Creating Psychological Safety in the Workplace for Learning, Innovation, and Growth. Hoboken, NJ: Wiley.
14. Goodhart, C. A. E. (1984). Monetary Theory and Practice: The UK Experience. Macmillan.
15. Heraclitus. Fragments, DK 22 B12. (English rendering used in this paper: Burnet, 1930; see entry above.)
16. Kaptein, M. (2008). Developing and testing a measure for the ethical culture of organizations: The corporate ethical virtues model. Journal of Organizational Behavior, 29(7), 923 to 947.
17. Keller, S. & Price, C. (2011). Beyond Performance: How Great Organizations Build Ultimate Competitive Advantage. Wiley. (McKinsey Organizational Health Index framework.)
18. MacIntyre, A. C. (2007 [1981]). After Virtue: A Study in Moral Theory (3rd ed.). Notre Dame, IN: University of Notre Dame Press.
19. McEvily, B., Perrone, V. & Zaheer, A. (2003). Trust as an organizing principle. Organization Science, 14(1), 91 to 103.
20. Meyer, J. W. & Rowan, B. (1977). Institutionalized organizations: Formal structure as myth and ceremony. American Journal of Sociology, 83(2), 340 to 363.
21. Schein, E. H. (2010). Organizational Culture and Leadership (4th ed.). Jossey-Bass.
22. Sull, D., Sull, C., & Chamberlain, A. (2020). When it comes to culture, does your company walk the talk? MIT Sloan Management Review (July 21, 2020).
23. Victor, B. & Cullen, J. B. (1988). The organizational bases of ethical work climates. Administrative Science Quarterly, 33(1), 101 to 125.
24. Watkins, M. D. (2003). The First 90 Days: Critical Success Strategies for New Leaders at All Levels. Harvard Business School Press.
25. Williamson, O. E. (1975). Markets and Hierarchies: Analysis and Antitrust Implications: A Study in the Economics of Internal Organization. New York: The Free Press.
26. Williamson, O. E. (1985). The Economic Institutions of Capitalism: Firms, Markets, Relational Contracting. New York: The Free Press.
27. World Management Survey. (n.d.). About the survey. Centre for Economic Performance, London School of Economics. Retrieved May 1, 2026, from https://worldmanagementsurvey.org
28. Zaheer, A., McEvily, B. & Perrone, V. (1998). Does trust matter? Exploring the effects of interorganizational and interpersonal trust on performance. Organization Science, 9(2), 141 to 159.
29. Anthropic. (2026). Claude [Large language model]. https://claude.ai
30. OpenAI. (2026). ChatGPT [Large language model]. https://chat.openai.com

---

Appendix H: Philosophical Foundations: Virtue Ethics from Aristotle to MacIntyre

VCAF's vocabulary borrows from a tradition older than management research. The framework's choice of "virtue" as its unit of measurement, rather than "value" or "competency" or "behavior," is deliberate and traces a specific theoretical lineage from Aristotle's Nicomachean Ethics through MacIntyre's modern recovery of virtue ethics. This appendix names the lineage and the work it does for the framework.

Aristotle: virtue as habituated excellence. The opening of Book II of the Nicomachean Ethics states that "moral virtue comes about as a result of habit... we become just by doing just acts, temperate by doing temperate acts, brave by doing brave acts." In Aristotle's account, virtue is a settled disposition produced through repeated action and shaped by the character of the community in which the action takes place. The distinction from belief or declaration is what makes virtue a workable category for VCAF. Virtue is read through observable behavior, since self-report misses what the gap between declaration and practice exposes. The gap itself is the measurement layer the framework operates on. Virtue is also acquired in context, which means firms produce or fail to produce virtue through the management systems that reward, punish, signal, and select on it. VCAF's Step 6 contradiction map runs on this Aristotelian premise.

MacIntyre: practices, internal goods, and the unity of a life. MacIntyre (2007 [1981]) recovers Aristotle's frame for the modern setting where shared moral vocabulary has fragmented and reduces, in his diagnosis, to expressions of preference rather than to grounded judgment. After Virtue supplies the conceptual machinery VCAF runs on.

A practice, in MacIntyre's sense, is "any coherent and complex form of socially established cooperative human activity through which goods internal to that form of activity are realized." Manufacturing, healthcare delivery, and financial intermediation organized around competence are all practices in this sense. Practices have internal goods, the satisfactions and capacities that come from doing the work well, and external goods, like compensation and reputation, which can be obtained through other paths. The internal goods of a practice cannot be acquired any other way. An organization that holds its declared virtues only as external claims forfeits the internal goods of the practices it conducts. VCAF measures that forfeit through the Pₖ score: how much of the declared virtue the firm's daily practice realizes.

The unity of a life is MacIntyre's argument that virtue is intelligible only across the entire life of an agent, never in isolated decisions. A firm whose declared virtues hold in moments of high visibility (annual reports, investor calls, hiring pages) but break down in the cadence of weekly reviews and forecast misses is the organizational analog of a life that lacks unity. The firm's MIₖ score is the gap between the virtue it claims and the virtue its decision arc traces.

What virtue ethics earns the framework. The choice of virtue rather than value as the unit of analysis carries weight. Values, in the modern management literature, are typically treated as preferences leadership selects and communicates. Virtues, in the Aristotelian and MacIntyrean traditions, are habituated capacities the organization either cultivates or fails to cultivate through its systems. A values framework can rest on declared statements; a virtue framework cannot. The Dₖ/Pₖ split is the framework's commitment to that distinction, and the EAI translation is the framework's claim that the gap has operating-profit consequences because organizations whose practiced behavior diverges from their declared character incur coordination costs the management literature has long named but not measured at virtue resolution.

The Heraclitus passage in Appendix D Step 3 belongs to the same tradition. "You cannot step twice into the same rivers" (DK 22 B12; trans. Burnet, 1930) names the temporal reality that an acquired firm is not the same firm a year into the operator's tenure. Virtue acquired across that arc is what produces the change.

---

Method Acknowledgement

The author used Claude (Anthropic) and ChatGPT (OpenAI) throughout the project for structural thinking, drafting, editing, and literature-review assistance. All intellectual claims, frameworks, analytical choices, and conclusions are the author's. The author reviewed, edited, and takes full responsibility for all content.

Submitted to the Koch Center for Family Enterprise, May 2026.