KeptScore Methodology 2.0

Public preview, open for comment. Published 10 October 2026.

Status. Kept Scores are private for now and will be public soon. This page sets out, in advance, exactly how they will be calculated and governed, so that companies, investors, researchers and public bodies can examine and challenge the method before a single score is published. The rules for selecting commitments and deciding outcomes are on the main methodology page (version 1.2), and remain in force.

What the KeptScore measures

One thing: how often an organisation delivers its own dated public commitments by the date it set itself. It is built only from what the organisation announced and what the public record later showed. It is not a credit rating, an ESG rating, an investment recommendation or a judgement of management’s integrity, and it does not predict the future.

Principles

The method is designed to meet the expectations regulators and standard-setters have set for ratings and data products, including the themes of IOSCO’s 2021 recommendations for ESG ratings and data providers: documented and transparent methodology, independence and management of conflicts of interest, fair engagement with the entities covered, and governance of methodology changes.

1. The organisation sets the bar. We never set targets. Every deadline is the organisation’s own, in its own words, with a link to the original document.
2. Silence never counts. Absence of evidence is not failure. A commitment counts against an organisation only when a public source shows it was late, partly delivered, not delivered or withdrawn.
3. Every number can be checked. Every commitment behind a score is listed with its sources, a permanent identifier and an archived copy, so anyone can recompute the score by hand.
4. Nobody can pay to change it. No company, subscriber or customer can pay to add, change, suppress or rank an entry, an outcome or a score.
5. Changes are public and dated. A score quoted today can always be understood later.

The calculation

• KeptScore (headline). Commitments delivered on time ÷ all commitments with a clear outcome whose deadline fell in the last 3 years, × 100. Met counts as on time. Late, Partly met, Not met and Withdrawn count as not on time. Superseded, Condition not met, No record found and disputed entries are left out.
• Minimum depth. No score until at least 15 outcomes in the window are confirmed.
• Confirmed outcomes only. An outcome moves a score only when a person has confirmed it, or independent checks agree on a primary source (an SEC filing, a regulator, an exchange or the organisation itself): two for a positive outcome, three for Not met or Withdrawn. Each score will show how many of its outcomes were confirmed by a person.
• Uncertainty shown. Each score will be published with its numerator and denominator, the share of outcomes with no public record, and a 95% interval. 12 met of 15 is 80%, but the interval is roughly 55%–93%; 80 of 100 is also 80%, with an interval of roughly 71%–87%. Bands will not imply more precision than the sample supports.
• Bands. 90+ Very strong, 75–89 Strong, 60–74 Mixed, 40–59 Weak, below 40 Very weak. The band is a label for the number, not a separate opinion.

New in version 2.0 (proposed)

Each item below is proposed for comment. None changes how outcomes are decided.

1. Every score shows its uncertainty

A score built on 15 outcomes is less certain than one built on 80. Each score will be published with the number of outcomes behind it and a 95% confidence range (Wilson interval), for example “78 (range 66–87, 41 outcomes)”. Two companies whose ranges overlap should not be described as different.

2. A long-run record alongside the headline score

The headline score uses the last 3 years. A separate KeptScore 10-Year Record covers every confirmed outcome since 2016, so readers can see whether recent performance is typical. The two are always labelled and never blended.

3. A difficulty-adjusted view

Some commitments are harder to deliver on time than others. In our record, announced transactions close on time far more often than new facilities open on time. An organisation that commits to harder things should not look worse simply for that. The Mix-adjusted KeptScore compares each organisation with what an average organisation would have achieved on the same mix of commitment types, and is shown next to the headline score, never in place of it.

4. Peer context

Each score shows the organisation’s position within its sector and index, so a reader can tell whether a number is high or low for that kind of business.

5. Execution signals, shown separately

Four measures sit beside the score and are never added into it: deadline drift (how far dates were moved before they fell due), scope drift (whether stated targets were reduced), language drift (whether firm wording softened to “expects” or “aims”) and average delay when late.

6. Point-in-time history

Each score is stored as it stood at every month end, using only information public by that date. A score is never silently rewritten: if a correction changes it, the earlier value, the change and the reason are kept.

How organisations can respond

• Notice before a first score. Before any organisation’s first score is published, it is notified at the investor-relations or general-counsel contact it publishes and given 10 business days to review the underlying commitments and outcomes.
• Correction of facts. Any factual error, such as a wrong deadline, a missed delivery announcement or a misread commitment, is corrected once shown, with the correction logged on the entry. The process is open to everyone on the same terms through the corrections page.
• Right of reply. An organisation may publish a short statement of context on any entry, such as a regulator’s delay, and it is shown next to the entry. A statement does not change an outcome; evidence does.
• Disputes. An entry under active challenge is labelled disputed and excluded from the score until it is resolved.

Accuracy, measured and published

The record is checked against a hand-reviewed random sample, and the results are published whatever they show. The first check, in September 2026, covered 100 published records:

Every error found was corrected, and the rules that produced it were tightened. Accuracy is re-measured on a fresh sample before scores are published and at least twice a year afterwards. Only confirmed outcomes ever count in a score.

Governance

• Versioned and dated. Every version of this methodology stays public, with the date it took effect.
• Consultation before change. Material changes are published for at least 30 days of public comment before they take effect, with a summary of the comments received and our response.
• Independent review. We are forming a Methodology Advisory Panel of independent specialists in securities disclosure, accounting research and public administration. Its members, and any of their interests, will be listed here.
• Separation of products. Commercial products such as data licences and research reports are sold to users of the data. They give no influence over any score, and the people who decide outcomes are not paid by results.
• Same rules for everyone. The same method applies to every organisation, whatever its size, sector or relationship with us.

Beyond listed companies

The method is designed to work for any organisation that makes dated public commitments: listed companies first, and in time suppliers, project developers and public bodies whose commitments are published. Each new setting will have its own published source rules and its own consultation before any score is issued.

Comment on this preview

We welcome comments from companies, investors, researchers, regulators and public bodies until 15 December 2026. Write to contact@keptindex.org with “KeptScore 2.0 comment” in the subject. Unless you ask otherwise, we may publish a summary of comments without naming you.

Version 2.0, public preview, 10 October 2026. Proposed. Version 1.2 remains in force until 2.0 is adopted after consultation. No Kept Score has been published under any version.

“Kept Score” and “KeptScore” are trademarks of Kept Index LLC (application pending).