An execution score methodology converts a public company's stated commitments into a measurable record of delivery. It identifies each claim, fixes its baseline and deadline, classifies the outcome from public records, and combines claim-level results into a company-level figure that can be reproduced by a second analyst. The term also names a fitness metric on consumer wearables; this article addresses the corporate-disclosure meaning, where the object of measurement is the distance between narrative and delivered results.
TL;DR
- An execution score measures delivery against specific, dated, verifiable claims, not financial performance and not disclosure quality.
- Every claim needs a recorded speaker, date, channel, exact wording, baseline, deadline and outcome source before it can be scored.
- Partial delivery, revised targets, delays, cancellations and unmeasurable claims each require an explicit scoring rule, stated in advance.
- Weighting by materiality and recency, plus a retained evidence trail, makes the score auditable and comparable across issuers.
- Uncertainty (heterogeneous voluntary disclosures, ambiguous claims) must be disclosed alongside the score, not absorbed into it.
Table of Contents
- What an execution score should measure
- Documenting a claim: baseline, deadline, outcome
- Scoring partial delivery, revisions, delays and unmeasurable claims
- Weighting and aggregation into a company-level score
- Making the score reproducible and auditable
- Bias and uncertainty: what must be disclosed
- Validating the score without circularity
- Perspective: why a standardized execution score matters
- Applying the methodology with Lacuna Index
- FAQ
- Sources
What an execution score should measure
The construct is narrow by design. An execution score answers one question: of the commitments a company made publicly, what proportion was delivered, on what timetable, and to what degree? It does not rate profitability, share-price return, or the eloquence of the disclosure. Conflating those quantities is the most common design error, because a company can deliver on modest promises while earning poor returns, or earn strong returns while abandoning its stated strategy.
The need for the construct is documented. Empirical work on the 2019 Business Roundtable statement found that, despite the public pledge, most signatory companies retained governance guidelines reflecting shareholder primacy, a divergence between corporate narrative and internal implementation (Bebchuk and Tallarita, 2022). A score that captures such divergence has to be anchored in what was said and what was subsequently observable, which leads to the following design requirements:
- Claim-level unit of analysis. The score aggregates individual commitments, never general impressions of a management team.
- Independence from valuation. Market reaction is a separate variable, used for validation and never as an input.
- Public-record evidence only. Inputs are earnings call transcripts, SEC filings, proxy statements, press releases and public statements, so that any reviewer can retrace them.
- Standardized rules across issuers. Identical definitions and outcome classes apply to every covered company, which is what permits cross-issuer comparison.
Readers interested in the promise-level building block can consult the guide on measuring CEO promise fulfillment rate, which treats the single-executive case in detail.
Documenting a claim: baseline, deadline, outcome
A statement becomes scorable only when it is converted into a structured record. Research on corporate promise verification, including the 2025 ML-Promise dataset, describes a sequence of promise identification, linkage to actionable evidence, an assessment of the clarity of each promise-evidence pair, and an inferred verification timeframe, for example within two years, two to five years, or longer. The same logic transfers directly to execution scoring.
| Record field | What it captures | Typical public source |
|---|---|---|
| Speaker and role | Who made the claim and with what authority | Call transcript, proxy statement |
| Date and channel | When and where the claim was made | Filing date, press release timestamp |
| Exact language | The original wording, unparaphrased | Transcript, filing text |
| Baseline | The starting value or condition | Prior-period filings |
| Target and deadline | The quantity and date promised, or an inferred horizon | Same statement, later restatements |
| Outcome evidence | The later observable result | Subsequent 10-K, 10-Q, 8-K |
| Clarity rating | Whether the claim is verifiable as stated | Analyst determination, logged |
The record structure mirrors good practice for disclosure disputes, where a document-backed chronology identifies the speaker, date, channel and exact original language and links each entry to available transcripts, filings and board materials.
Which filings carry commitments
Forward-looking commitments are not confined to executive remarks. Under Regulation S-K, Item 303, the MD&A section must address known trends, demands, commitments, events and uncertainties reasonably likely to have a material effect, as set out in SEC Release No. 33-8350. That guidance also indicates that forward-looking information is required in MD&A in some circumstances, particularly for known material trends and uncertainties. For the analyst, MD&A is therefore a regulated location in which management's own expectations are stated and later testable, and it generally carries more evidentiary weight than an unsworn remark on a call.
Disclosure of this kind also frames what investors should be able to expect from management more broadly; an investor due diligence checklist illustrates how stated commitments and risks are expected to appear in material financial reporting. The related article on how public disclosures reveal management gaps shows the same documents read for inconsistency.
Scoring partial delivery, revisions, delays and unmeasurable claims
The decisive methodological choices concern outcomes that are neither clean successes nor clean failures. Public-sector performance frameworks offer tested precedents, and several can be adapted to corporate claims.
The California Energy Commission's evaluation approach deducts points for avoidable due-date changes or late submissions, distinguishes changes outside the recipient's control, and rates partial completion, withdrawal, and cancellation or termination separately (December 2021). The federal Supplier Performance Risk System goes further on delivery: it scores a weighted combination of the percentage of line items delivered on time and the average days late, treats early and on-time deliveries as positive, and treats late, terminated and open deliveries as negative, while excluding termination for convenience from delivery calculations (Version 3.3, March 2023).
Adapted to corporate commitments, a defensible outcome taxonomy is the following.
| Outcome class | Definition | Scoring treatment |
|---|---|---|
| Delivered on time | Target met by the stated deadline | Full credit |
| Delivered early | Target met ahead of deadline | Full credit, optionally a small timeliness increment |
| Delivered late | Target met after deadline | Partial credit, reduced by delay length |
| Partially delivered | Part of the quantified target met | Credit proportional to the share delivered |
| Revised | Target or deadline changed after the fact | Scored against the original claim; the revision is logged with its stated rationale |
| Withdrawn or cancelled | Commitment abandoned or discontinued | Negative, distinguished by whether circumstances were external |
| Open past deadline | No outcome evidence after the horizon | Negative until evidence appears |
| Not verifiable | No measurable outcome could ever be observed | Excluded from the numerator, counted in the clarity ledger |
Two principles govern this table. First, revisions must not erase the original claim: if a target moved, the score reflects the commitment as first made, with the change noted, because otherwise a company can improve its record by rewriting its goals. Second, the treatment of externally caused changes must be a stated rule rather than an analyst's discretion. The California precedent separates avoidable from unavoidable changes, and corporate scoring needs an equivalent test (for example, a documented regulatory action or a force majeure event disclosed in a filing).
Handling claims with no measurable outcome
Vague claims are a finding, not a gap in the data. In the ML-Promise framework, clarity combines the quantity and quality of evidence, and evidence resting on unproven carbon-capture technology is given as an example of a claim classified as not clear. A rigorous execution score therefore keeps unverifiable claims out of the delivery ratio but reports their share separately. A company whose commitments are mostly unfalsifiable earns no flattering score; it earns a high disclosure-opacity reading, visible beside the score.
Weighting and aggregation into a company-level score
Claim-level results must be combined according to rules that reflect materiality, time and sector context.
Materiality. A commitment tied to a core revenue line or a capital program should weigh more than an aspirational remark about a peripheral initiative. Materiality tiers should be assigned from the disclosure itself, for instance by whether the claim appears in MD&A, in guidance, or only in a conference remark, and recorded in the evidence log.
Recency. SPRS offers a working model: it age-weights performance data daily, so that newer data contributes more. The same source gives deliveries with no associated quality record a +0.1 positive credit after 90 days, an example of how a neutral absence of adverse evidence can be given explicit, bounded treatment rather than left implicit. A corporate score may adopt decay weighting so that a restructured management team is not permanently bound to commitments from a distant period, while retaining the older record for trend analysis.
Horizon. Claims maturing within a year and claims maturing in five years should not be pooled without adjustment. Open claims inside their horizon are pending, not failed, and the open-past-deadline class above applies only once the horizon has lapsed.
Sector context. Raw scores are most informative against peers. Disclosure norms differ between, for example, health care, industrials and financials, so a company's score is read against its sector distribution. Sector benchmarks also expose structural differences, such as long capital-project horizons in industrials, that would otherwise be mistaken for poor execution.
The board-level counterpart of this aggregation is described in measuring the board promise-delivery gap for investors.
Making the score reproducible and auditable
A score that cannot be rebuilt by a second analyst is an opinion. Audit evidence standards provide a mature vocabulary for the requirement. Under PCAOB standards on audit evidence (AS 1105), auditors must obtain sufficient, appropriate evidence that is both relevant and reliable, and evidence from independent third-party sources is generally considered more reliable than evidence from internal company sources. The 2024 amendments add that, for electronic information received from external sources, the auditor must understand its source and the company's receipt, maintenance and processing, and test the information or related controls.
An execution methodology is not an audit, and a public-records analyst has no access to a company's internal controls. The principle of an evidence hierarchy nevertheless transfers:
- Outcome evidence from filed, independently attested documents (audited financial statements, regulator-received filings) outranks outcome evidence from press releases.
- Original text outranks summaries. Transcripts and filings are cited verbatim; secondary coverage is not used to establish an outcome.
- Each score element links to a source passage. A qualified reviewer should be able to reconstruct material entries from specific source passages and reproduce the calculations.
For workflows that use automated extraction, the standard of proof is stricter than a saved output. Guidance on finance workpapers holds that a transcript or AI-generated record is insufficient proof of reproducibility; an audit-ready trail shows the reconstruction of the run, the recalculation of material figures, and the reviewer's specific inspection, challenge and conclusion. In methodological terms, every machine-extracted claim should be confirmed by a human reviewer whose determination is logged, and the governance log should record changes to rules and reclassifications over time.
Cross-issuer comparability follows from the same discipline: one rulebook, one outcome taxonomy, one weighting scheme, applied to all issuers without proprietary data or insider access. The methodology page on Lacuna Index's methodology walkthrough sets out how this process is organized.
Bias and uncertainty: what must be disclosed
No execution score is free of uncertainty, and a credible methodology states where it enters.
- Heterogeneous disclosure. Academic work notes that voluntary disclosures are heterogeneous, inconsistent and often qualitative, and may be less reliable than mandatory disclosures, and that validating whether a metric captures strategic bias or intentional error is challenging. Scores should flag which claims rest on voluntary versus mandatory sources.
- Restricted evidence. Where third-party evidence carries restrictions, limitations or disclaimers, PCAOB standards require the auditor to evaluate the effect of those restrictions on reliability. An analyst should apply an analogous reading to outcome statements qualified by disclaimers.
- Claim-selection bias. Companies choose which commitments to publicize. A score built only from prominent claims inherits that selection; documenting the extraction universe (which documents, which period) limits the effect.
- Classification judgment. Clarity ratings and materiality tiers involve analyst judgment. Inter-reviewer agreement and logged overrides are the appropriate disclosure.
- Pending outcomes. The share of claims still inside their horizon should be reported, since it bounds how much the score can still move.
The companion piece on corporate promise gap governance risk discusses how these uncertainties bear on governance decisions.
Validating the score without circularity
Validation asks whether the score predicts or corresponds to something independent of its construction, and the main hazard is circularity: if financial results feed the score, correlation with financial results is guaranteed and uninformative.
A useful template exists in accounting research on non-GAAP exclusions. The study by Dechow, Loh and Wang compared its ratings with regulator guidance, CFA Institute survey results and the judgments of academics and analysts, and then tested associations with later regulatory scrutiny, analyst forecast dispersion and investor price discovery (July 2023 draft). The two-stage design is instructive: convergent validity against expert judgment first, predictive validity against later observable events second.
For execution scoring, the corresponding tests include:
- Convergent checks against independent expert classifications of a sample of claims.
- Predictive checks of whether low execution scores precede later restatements, guidance cuts, governance actions or elevated forecast dispersion.
- Discriminant checks that the score is not simply a proxy for size, sector or disclosure volume.
- Hold-out periods, so that rules are fixed before outcomes are examined.
The investment application of validated scores is covered in the public disclosure gap investment analysis guide, and journalistic use in corporate promise tracking journalism methods.
Perspective: why a standardized execution score matters
We built the execution score on a single premise: a claim that cannot be dated, baselined and checked against a later public record is not evidence of anything, and a market that prices such claims is pricing narrative.
The strongest argument for a standardized score is comparability. Qualitative assessments of corporate accountability tend to differ by author, by sector and by year, which makes them difficult to audit. A single rulebook, applied to every covered issuer and recorded in an explicit governance log, allows a reviewer to disagree with a specific classification rather than with an overall impression.
We also hold that public records are sufficient. Earnings calls, SEC filings, proxy statements and press releases already contain the commitments and, over time, the outcomes. The analytical work lies in structuring them consistently, which is why we classify companies by the relationship between delivery and valuation (earned, borrowed, undervalued) instead of by sentiment.
For institutional investors, governance advocates, proxy advisors and financial journalists, the practical value is a score that can be challenged, reproduced and compared across a cohort.
Glen
Applying the methodology with Lacuna Index
Lacuna Index applies this process to public companies using public records only, producing execution and narrative scores, sector benchmarks, performance archetypes and audit-traceable scoring. Sector benchmarks are available without a subscription, for instance financial services and insurance, while per-company forensic analysis is gated for subscribers.
Institutional teams evaluating coverage can review the Cohort Pulse and Sector Sweep plans or the per-company analysis pages for current terms.
FAQ
Is an execution score the same as a financial performance score?
No. An execution score measures delivery against specific public commitments, regardless of whether the results were profitable. Financial performance and market returns are treated as separate variables and used only to validate the score against later outcomes, which avoids circular reasoning.
How are revised or withdrawn commitments scored?
The original commitment remains the reference point, and the revision is logged with the company's stated reason. Public-sector evaluation frameworks separate avoidable changes from those outside the actor's control and rate withdrawal and cancellation distinctly, and corporate scoring can follow the same logic with a rule fixed in advance.
What happens to claims that cannot be measured?
They are excluded from the delivery ratio and reported separately as a clarity or opacity indicator. A company with many unverifiable claims does not gain a favorable score from them, but the pattern itself becomes visible to the reader.
Can an execution score be reproduced by another analyst?
Yes, provided each score element links to a source passage, the outcome taxonomy and weights are fixed, and reviewer determinations are logged. Without those elements, a second analyst cannot reconstruct the figure, and the score should not be considered auditable.
Sources
- Have Business Roundtable Companies Lived Up to Their Commitments (ProMarket, 2022)
- SEC Release No. 33-8350, Commission Guidance Regarding MD&A
- ML-Promise: A Multilingual Dataset for Corporate Promise Verification (EMNLP 2025)
- ML-Promise (arXiv)
- California Energy Commission, Grant Recipient and Contractor Past Performance Evaluation
- SPRS Evaluation Criteria Manual, Version 3.3
- PCAOB Auditing Standards, AS 1105
- PCAOB amendments to AS 1105, Audit Evidence (2024)
- AI Audit Trails in Finance: 6 Workpaper Questions
- How to build a document-backed case chronology for securities disclosure disputes
- Beyond earnings quality: evaluating the quality of voluntary disclosures
- Dechow, Loh and Wang, non-GAAP exclusions (July 2023 draft)
- Investor Due Diligence Checklist