Governance Analytics Methodology: How Scores Are Built

A governance analytics methodology is the documented process by which public evidence about a company is converted into comparable, auditable measures of accountability. The term covers two distinct things that search results often conflate: the governance of an organization's own data and analytics, and the analysis of a public company's governance and accountability. This article addresses the second meaning: how indicators are selected, how scores are constructed, how reliability is tested, and how the results are used without overstating what they show.

TL;DR

• Governance analytics for public companies converts disclosures (filings, proxy statements, earnings calls, press releases) into standardized indicators that can be compared across issuers.
• Established scoring systems share a pattern: many observable indicators, grouped into dimensions, standardized, then aggregated into a relative score.
• A methodology is credible only if it can be tested for validity and reliability and if every score traces back to a citable record.
• Disclosures are partial by construction, so every score is a point-in-time estimate that needs uncertainty reported alongside it.
• Lacuna Index applies this logic to the gap between narrative claims and delivery, using public records only.

Table of Contents

• 1. What governance analytics measures
• 2. Which public records feed the analysis
• 3. Identifying claims and outcomes in the text
• 4. How indicators, scores and benchmarks are constructed
• 5. Testing reliability, validity and reproducibility
• 6. Biases and limitations of disclosure-based evidence
• 7. Using results responsibly by audience
• 8. Our perspective
• Applying the method with Lacuna Index
• FAQ
• Sources

1. What governance analytics measures

At the issuer level, governance analytics measures whether the structures and conduct of a company support accountability to its owners and the public. Conventional models concentrate on structure. ISS's governance QualityScore, for example, assesses governance-related risk across four pillars: audit and risk oversight, board structure, compensation, and shareholder rights and takeover defenses. These are inputs to accountability: the machinery that should, in principle, constrain management.

A second family of methods measures outcomes and conduct. Rather than asking whether a board has the expected architecture, such methods ask whether the company did what it said it would do. This is the approach taken by the aspiration-to-execution framework, which treats a public commitment as a testable claim and subsequent reporting as the evidence of delivery. The two families are complementary: structural scores describe capacity for oversight, while delivery scores describe the observable record.

The distinction matters for method selection. A structural model can be built from checklists and is comparatively cheap to reproduce. A delivery model requires reading unstructured text, identifying what was promised, assigning a time horizon, and locating the later disclosure that confirms or contradicts it. The remainder of this article describes the second type in the most detail, while drawing on published structural methodologies for the mechanics of scoring.

2. Which public records feed the analysis

A methodology based solely on public records draws on a bounded set of document types, each with a distinct function and a distinct bias.

The regulatory frame explains why these documents are usable as evidence. Management's discussion and analysis is governed by Item 303 of Regulation S-K, which the SEC amended in 2020 to modernize and streamline the rules. Selective communication is constrained by Regulation Fair Disclosure, which requires public companies to make material nonpublic information available broadly when they share it with certain market professionals or shareholders. The consequence for method design is practical: material statements tend to migrate into public channels, which makes a public-records-only approach defensible without insider access.

For readers who want the validation side of this step, how to validate corporate claims in public disclosures sets out the cross-checking sequence in detail.

3. Identifying claims and outcomes in the text

Claim extraction is the least standardized stage in any disclosure-based method and therefore the one that most needs written rules. A workable protocol defines a claim as a statement that is specific, attributable to an authorized speaker, and capable of being confirmed or contradicted by a later public record.

1. Segment. Split each document into discrete statements and discard purely descriptive or historical language.
2. Classify. Tag each remaining statement by type: financial target, operational milestone, governance commitment, or strategic positioning.
3. Anchor. Record the speaker, the document, the date and the stated horizon, so the claim can be located again.
4. Match. Link the claim to the later disclosure that reports the outcome, which may be a subsequent filing, a restated target, or an unexplained silence.
5. Grade. Assign a delivery status (delivered, partly delivered, not delivered, withdrawn without explanation) using fixed criteria.

The fifth step produces the quantity that distinguishes delivery analytics from structural scoring: the gap between what was claimed and what was reported. Silence is itself a data point. A target that quietly disappears from later materials is graded differently from one that is explicitly revised with a reason. Patterns of this kind are catalogued in governance red flags in public disclosures, and the underlying concept of disclosure integrity is covered in corporate disclosure integrity explained for analysts.

4. How indicators, scores and benchmarks are constructed

Published methodologies converge on a recognizable construction. Examining two of them clarifies which design choices are structural and which are discretionary.

Indicators grouped into dimensions

The Drucker Institute's 2024 U.S. company-effectiveness methodology uses 35 indicators across five dimensions: customer satisfaction, employee engagement and development, innovation, social responsibility, and financial strength. It treats effectiveness as a latent construct, meaning a quantity that cannot be observed directly and is inferred from several observable measures. The same logic applies to accountability: no single filing proves it, but a consistent pattern across many indicators can support an inference.

Standardization and aggregation

The Drucker procedure standardizes indicator data, averages indicators within each dimension, restandardizes the dimension scores, converts them to T-scores, and applies factor analysis to derive the overall result. Missing values are handled using the valid indicators available within the relevant dimension. Three design choices are visible here and each is a point at which a reviewer should ask for documentation: how indicators are weighted, how absent data is treated, and which transformation produces the final scale.

Relative versus absolute scales

Governance risk models frequently report scores relative to a peer population. An independent explanation of QualityScore reports that the model evaluates more than 230 factors against market best practices and converts raw results to scores from 1 (lowest risk) to 10 (highest risk), relative to other U.S. companies. A relative scale is useful for ranking, but it moves when the peer group moves: a company's score can change with no change in its own conduct. Methodologies should therefore state the comparison universe and the date on which it was fixed.

Archetypes and benchmarks

A delivery-focused methodology adds a classification layer. Lacuna Index pairs an execution score (how much of the claimed agenda was delivered) with valuation context, and sorts companies into archetypes such as earned, borrowed or undervalued. An earned position combines strong delivery with a valuation that reflects it. A borrowed position reflects a valuation that rests on narrative more than on delivery. An undervalued position indicates delivery that the market price does not appear to reflect. Sector benchmarks then aggregate company-level results so that an individual score can be read against a defined cohort. The principles behind that aggregation are described in public company accountability benchmarks, and the use of scores in decisions in accountability scoring in governance decision making.

5. Testing reliability, validity and reproducibility

A score that cannot be tested is an opinion with a number attached. Three properties should be demonstrated before a methodology is relied upon.

• Validity: the indicators measure the construct they claim to measure. The Drucker Institute tested construct validity, reliability and goodness of fit using structural equation modeling, which is one accepted route for latent constructs.
• Robustness to missing data: results should not depend on the method used to fill gaps. The same Drucker documentation compares its missing-data treatment with multiple imputation, a direct test of that sensitivity.
• Reproducibility: a second analyst applying the same written rules to the same records should reach the same grades. For text-based methods this is tested through independent double coding of a sample of claims and reporting the rate of disagreement.

A fourth property is specific to evidence-citing systems: traceability. Every score should resolve to a list of cited passages, each with document, date and grading rationale, so that a reviewer can disagree with a single grade without discarding the whole result. Lacuna Index publishes governance logs for its scoring for this reason, and its methodology walkthrough documents the sequence from record to score.

6. Biases and limitations of disclosure-based evidence

Disclosure is not a neutral record. Westbrook's 2004 analysis of the Sarbanes-Oxley Act argues that corporate disclosure is reductive and subjective: companies frame their own narrative, so every disclosure is partial and shaped by the discloser, even when made in good faith. Any methodology built on disclosures inherits that property.

The principal biases and the corresponding controls are as follows.

The Drucker Institute offers a useful framing for uncertainty: it describes its ranking as a point-in-time snapshot rather than proof of causation. The same restraint applies to delivery scores. A low execution score documents a pattern in the public record. It does not establish intent, and it does not by itself establish misconduct. Reports should therefore carry the evidence cutoff, the number of claims graded, the share of claims excluded as unfalsifiable, and any indicators for which data was absent.

7. Using results responsibly by audience

Different users act on the same score in different ways, and each carries a distinct obligation.

• Institutional investors and research analysts: treat the score as a screen for questions, not a conclusion. The gap between narrative and delivery identifies where to seek further evidence; public disclosure gap investment analysis outlines how that is done.
• Governance advocates and proxy advisors: cite the underlying records rather than the aggregate score, so that engagement with a board rests on verifiable passages. The wider consequences of misstatement are discussed in corporate misrepresentation and governance impact.
• Financial journalists: report the evidence and its date, name the limits of inference, and give the company the opportunity to respond. Possible angles are collected in public company accountability story angles.
• Regulators: use delivery gaps for prioritizing review, while recognizing that a score is not a finding under any legal standard.

Underlying all of these is the broader premise that transparency supports governance, a point developed in why corporate transparency matters for governance.

8. Our perspective

We built Lacuna Index on one conviction: a governance assessment is only as useful as it is checkable. Structural scores tell a reader whether a board has the expected architecture. They do not tell a reader whether management did what it said. That gap, between the narrative a company constructs and the delivery the record shows, is where mispricing and weak oversight tend to hide, and it is the quantity we set out to measure.

Three commitments follow from that. First, we use public records only, with no proprietary data and no insider access, because a method that cannot be reproduced from public documents cannot be audited by those who rely on it. Second, we apply one standardized methodology to every covered company, because cross-issuer comparability disappears the moment rules are adjusted case by case. Third, we publish governance logs for scoring, so that a disputed grade can be traced to a specific passage and challenged on its merits.

We also hold that a score should never be read as a verdict. Disclosures are partial, and the output of any disclosure-based method is a documented estimate at a point in time. Our reports are designed to be argued with, and that is the standard we recommend applying to any governance analytics methodology, including ours.

— Glen

Applying the method with Lacuna Index

Sector benchmarks at Lacuna Index are free to read, starting with the Financials and Health Care cohorts, and they show how the aggregation described above looks in practice. Company-level forensic reports, with execution and narrative scores and the supporting evidence, are available through subscription: the Cohort Pulse and Sector Sweep plans cover cohort and sector work, and per-company analysis covers individual issuers. The user guide explains how to read a report.

FAQ

What is the difference between data governance and governance analytics?

Data governance concerns how an organization manages the quality, security and availability of its own data. Governance analytics, as used here, evaluates a public company's accountability from outside, using its published disclosures. The first is an internal control discipline; the second is an external assessment method.

What are the main pillars of governance that analytics models score?

Structural models commonly organize indicators into pillars. ISS's governance QualityScore uses four: audit and risk oversight, board structure, compensation, and shareholder rights and takeover defenses. Delivery-based models add a fifth lens, which is whether stated commitments were met.

Can a governance score be reproduced independently?

It can, provided the methodology documents its indicators, its grading rules and its evidence for each score. Reproducibility is tested by having a second analyst grade the same sample of records and reporting the rate of disagreement. A score without cited evidence cannot be reproduced and should be treated with caution.

Does a low execution score mean a company misled investors?

No. A low execution score shows that stated commitments were not matched by later public reporting. It does not establish intent or legal wrongdoing, and the explanation may be changed circumstances, a revised strategy or poor disclosure practice. The underlying evidence should be reviewed before drawing any conclusion.

Sources

• ISS Releases Updates to Governance QualityScore Model (PDF)
• All About Your Company's QualityScore Rating, Weaver
• Methodology 2024, The Drucker Institute
• Telling All: The Sarbanes-Oxley Act and the Ideal of Disclosure, Westbrook (2004)
• SEC Amends MD&A Disclosure Rules and Trims Financial Disclosure Requirements, Proskauer
• Regulation Fair Disclosure (FD), Cornell Legal Information Institute

Recommended

• Accountability Scoring in Governance Decision Making
• Benefits of Systematic Accountability Scoring for Investors
• The Undated Citation: Why a Governance Score Can Look Current While Its Evidence Is Two Years Old
• Audit Scoring for Managers: Freeze Inputs, Trace Every Result