The Duration Measurement Standard defines how the time a regulated outcome takes is measured and what a duration figure must carry. It defines the seven clocks, time at risk and its companion measures, and the start and stop events, censoring treatment, cutoff discipline and cohort rules that make a figure comparable. It requires that a duration figure be a band with a visible tail, stated against a named cohort and a named cutoff date.
Criterica Intelligence Research (2026). Duration Measurement Standard, Version 1.0-rc. Criterica. https://criterica.com/standards/duration-measurement. DOI: pending
@techreport{criterica2026durationmeasurement,
author = {{Criterica Intelligence Research}},
title = {Duration Measurement Standard},
institution = {Criterica},
type = {STD-003},
number = {STD-003},
year = {2026},
note = {Version 1.0-rc. DOI pending},
url = {https://criterica.com/standards/duration-measurement}
}1. Scope
This standard defines how a duration figure for a regulated outcome is measured and what that figure must carry when it is published or relied on. A regulated outcome is a matter decided by a court, a regulator or a contract: a lawsuit, an enforcement action, an insured claim, an arbitration.
It applies to any party that states how long a matter or a book of matters will take: a funder pricing an advance, an insurer setting a reserve, a firm planning its working capital, a data provider publishing an estimate. It covers the measurement of time. It does not cover who prevails, what a matter is worth or what a position will return. Those are separate questions with separate standards.
A duration figure is a distribution with a stated construction. It is never a single date. A statement that conforms to this standard reports a band, names its cohort and its cutoff, and says how open matters were treated.
2. Definitions
2.1 The seven clocks
A matter does not run on one clock. Seven distinct clocks sit inside every duration estimate, each with its own start event, its own end event, its own drivers and its own evidence base.
| Clock | Measures |
|---|---|
| Next procedural event | Time to the next consequential filing, ruling or hearing. The shortest clock, and the one that re-times the others. |
| Settlement | Time to a negotiated resolution. |
| Adjudication | Time to a decision on the merits. It runs on the court’s calendar, not the parties’. |
| Appeal | A discrete additional layer with its own distribution, conditional on the route taken to it. |
| Enforcement | Time from judgment to enforceability. |
| Cash collection | Time from enforceability to money received. Legal resolution and cash resolution are different dates. |
| Capital recycling | Time from commitment to redeployment. The clock a portfolio compounds on. |
The first five clocks are not sequential stages. Settlement, adjudication and dismissal race one another, and whichever endpoint arrives first decides both the duration and which later clocks ever start. A single expected duration averages across races with different finish lines. The proper object is a set of conditional distributions that update as events eliminate competitors.
2.2 Time at risk and its companion measures
| Term | Definition |
|---|---|
| Time at risk (TaR) | The family of tail-first measures of how long capital or an obligation stays exposed to an unresolved matter. |
| TaR90 | The 90th-percentile time to resolution. The tail a reserve must carry, stated as a number rather than discovered later. |
| Excess TaR | TaR90 minus the median. The distance between the figure a plan is built on and the figure one matter in ten actually delivers. |
| Duration drift | Movement of the duration distribution since underwriting or the last review, preserved against the original estimate. |
| Cash-at-Date | The probability that a defined share of a book resolves by a given period. |
| Delay concentration | The share of a book’s time exposure that rides a common driver, such as one court’s calendar, one doctrine or one procedural regime. |
| Recoverable time | The portion of expected delay that feasible action could remove without damaging position. Defined so that it can be measured. It is reported only where intervention histories and outcomes exist to support it. |
2.3 Terms used in the measurement rules
- —Checkpoint: an event that changes what is known about a matter and therefore re-times one or more clocks. Examples are a survived motion, a certification decision and a scheduling order.
- —Censoring: the condition of a matter that has not yet reached its end event at the time of measurement. Its true duration is at least its current age and is not otherwise known.
- —Cohort: the set of comparable matters a distribution is computed from, defined by court or forum, case family, filing window and any other stated filter.
- —Cutoff date: the date after which no information is used. Every figure is as of its cutoff date.
- —Band: a stated interval of a distribution, such as the interquartile range, reported with its construction.
3. Measurement rules
3.1 Start and stop events
Every duration figure names the clock it measures, the event that starts it and the event that stops it. The start and stop events are written down before the figure is computed and do not change to fit the result.
- 1.A figure built from the public record starts at filing and stops at the recorded termination of the matter. The figure says so.
- 2.A figure for capital starts at commitment and stops at the receipt of cash or at redeployment, depending on the clock. It does not borrow the court’s dates.
- 3.Where clocks race, the figure is conditional: it states which endpoint it assumes, such as if the matter settles or if the matter is adjudicated.
- 4.Where the record codes how a matter ended only coarsely, the figure does not infer more than the record supports. In particular it does not infer which negotiated outcomes were settlements or which party prevailed.
3.2 Censoring
Open matters have not finished. A method that quietly drops them is biased toward optimism, because the matters that remain are the ones that resolved quickly. A conforming figure treats censoring explicitly, using a method designed for it, and states the method. Where a figure is computed only from resolved matters, it says so, and it says that recent filing years thin the long-duration cells.
3.3 Cutoff and frozen-date discipline
A figure is computed on the information that existed at its cutoff date and on nothing later. A figure restated after the cutoff is a new figure with a new cutoff. The original is kept.
Estimates that are validated are validated on chronological splits: trained on an earlier period and tested on a later period the method never saw. Random splits hide regime change and are not used to support a duration claim.
3.4 Cohort definition
- —The cohort is defined by stated filters and the definition is published with the figure.
- —A conditional figure restricts the cohort to matters still open at the age of the matter being measured, so that the band reflects what remained, not what began.
- —Below a stated minimum cohort size the method returns a tier or a refusal, not a precise-looking band. The minimum is part of the method and is published.
- —Pooling across forums or case families is declared, with the reason.
3.5 Re-estimation and preservation
A duration estimate updates when evidence arrives, at each checkpoint, not when a review cycle happens to come around. Each prior band is preserved beside its successor. Preservation is what makes accuracy auditable after the fact and what makes duration drift measurable.
4. Reporting requirements
A duration figure that claims conformance carries each item in the table below. A figure missing an item is reported as non-conforming.
| Item | Requirement |
|---|---|
| Clock | Which of the seven clocks the figure measures, with its start and stop events. |
| Cohort | The filters that define the cohort, the filing window, and the forum or forums. |
| Cutoff date | The date after which no information was used. |
| Percentile basis | The percentiles reported, such as the median, the interquartile range and the 90th percentile, and whether they are conditional on age. |
| Censoring treatment | How open matters were handled and the method used. |
| Sufficiency | The minimum cohort size applied and whether the figure met it. |
| Validation | Whether the estimate was validated on a later period, and the period. |
| Band, not date | The figure is a band with a visible tail. It is not a single date. |
| Prior estimates | Where the figure replaces an earlier one, the earlier one is cited and kept. |
A duration figure that is attached to a return must state the cash-flow assumption behind the conversion. Duration changes the annualized return at a fixed multiple, and a multiple on its own is blind to time.
4.1 Averages and single dates
A mean duration is not reported as the duration of a matter or a book. Durations are right-skewed, and the mean hides the tail that a reserve, a facility or a refinancing has to carry. The median and the tail are reported together, and the gap between them is reported as excess time at risk.
A single date is false precision. A party that must give one date for planning gives the percentile it stands behind and the band around it, and says which clock the date belongs to. A legal resolution date is not a cash date, and a cash date is not a redeployment date.
4.2 Book-level figures
A book-level figure states how the matters were weighted, whether matters of different ages were pooled, and how much of the book’s time exposure depends on a common driver. Two books with the same median can carry very different time at risk, and only a tail-first measure distinguishes them.
5. Conformance statement
A party may state that a duration figure conforms to this standard only if the figure meets sections 3 and 4 in full. Partial conformance is not claimed. A conforming figure carries this statement:
Conforms to the Criterica Duration Measurement Standard, version 1.0-rc. Clock, cohort, cutoff, percentile basis and censoring treatment are stated above.
Conformance describes how a figure was measured. It is not an opinion on the merits of any matter, a forecast of any individual outcome or a statement about what a position will return.
6. Versioning and errata
This standard is a release candidate. It is open for comment until the next version is issued. Comments are read by Criterica Intelligence Research and answered in the version history.
- —A major version changes a requirement. Figures stated against an earlier major version say which one.
- —A minor version clarifies without changing a requirement.
- —An erratum corrects an error in a published version. Each erratum is dated, described and listed on the errata page for that version. The text of the published version is not silently edited.
7. References
- 1.Criterica Intelligence Research. The Seven Clocks: Why Litigation Duration Is Not One Number. Criterica Research. critericaintelligence.com/research/seven-clocks
- 2.Criterica Intelligence Research. Time-at-Risk: A Metric Family for Legal-Asset Duration. Criterica Research. critericaintelligence.com/research/time-at-risk
- 3.Criterica Intelligence Research. Duration and Returns: The Arithmetic of Time in Litigation Finance. Criterica Research. critericaintelligence.com/research/duration-and-returns
Tell us what should change in this standard.
Comment on a standard
Name the standard (STD-003), the section and the change you propose. Comments are read by Criterica Intelligence Research and answered in the version history.
Questions
Each standard page carries a plain citation and a BibTeX entry with a copy button. Cite the code, the title and the version. A release candidate is cited as a release candidate.
A standard marked request for comment is published in full and open to comment before it is issued as a numbered version. Comments are read by Criterica Intelligence Research and answered in the version history.
An error in a published version is recorded as a dated erratum on the errata page for that version. The published text is not silently edited.