Standards crosswalk: where the delay-analysis documents disagree
Four documents govern most schedule and delay work in the English-speaking world. They do not agree. This page sets out where they diverge, states each one's position, and stops there — it does not say which is right, because that is a question for the contract, the tribunal or the reviewing authority, not for a comparison table.
The gap this fills is real. The best existing side-by-side comparison of these documents is a conference paper held inside a member learning system, and no free four-way comparison that includes GAO-16-89G could be found.
The four documents
| Short name | Full title | Kind of instrument | Where it binds |
|---|---|---|---|
| AACE 29R-03 | AACE International Recommended Practice 29R-03, Forensic Schedule Analysis | Recommended practice — a taxonomy of methods | By adoption, or where a tribunal finds it authoritative. Not binding by its own force |
| SCL Protocol | Delay and Disruption Protocol, Society of Construction Law, 2nd edition (February 2017) | Guidance. Non-contractual on its own terms unless a contract adopts it | By adoption, largely UK and Commonwealth practice |
| ASCE 67-17 | ANSI/ASCE/CI 67-17, Schedule Delay Analysis | ANSI-accredited consensus standard | By adoption. On US federal work, UFGS 01 32 01.00 10 §3.8.4 makes it govern over AACE 29R-03 where a methodology chosen from 29R-03 conflicts with it |
| GAO-16-89G | GAO Schedule Assessment Guide: Best Practices for Project Schedules, December 2015 | Audit criteria for programme schedules | The federal audit community; widely used as a reference elsewhere |
One structural point before the table. The first three are disputes documents: they exist to say how a delay is analysed after the fact. GAO-16-89G is a programme-assessment instrument: it exists to say whether a schedule is a reliable forecasting tool while the work is still running. That is why it is usually missing from comparisons of this kind, and it is also why its omission costs something — two of the questions below are questions GAO answers directly.
The crosswalk
One row per question. Where a document says nothing on a question, that is recorded as saying nothing, which is itself a position.
| Question | AACE 29R-03 | SCL Protocol (2nd ed.) | ASCE 67-17 | GAO-16-89G |
|---|---|---|---|---|
| What is the critical path? | RP 49R-06 gives longest path as the definition, then documents four non-equivalent operational methods and declines to endorse one | Uses the longest-path conception; treats the choice of method as a matter for the facts | Longest-path framing, with compensability reasoning tied to the longest path | Practice 6: the path of longest duration through the sequence of activities |
| Which concurrency theory applies? | §4.2.D.1 sets out the literal and the functional theories and endorses neither | Core Principle 10 prescribes an outcome — effective cause, not mere overlap in occurrence — and its own ¶10.10 flags that position as provisional | Prefers apportionment where the facts allow it (§8.3), with an excusable-but-non-compensable fallback (§8.2) | Silent — not a disputes document |
| What does a concurrency finding do to an extension of time? | §1.5.B(6) requires the delay to affect the critical path; says nothing about the effect of a concurrency finding on the extension | CP10: a contractor's concurrent delay does not reduce the extension otherwise due | §8.3 prefers apportionment; §8.2 gives shared, non-compensable treatment where apportionment cannot be done | Silent |
| Must an intention to pace be noticed contemporaneously? | §4.2.G records that such notices are very unusual in practice and warns against treating one as a strict proof requirement | ¶15.2 recommends that notice be given | Does not set a notice precondition of this kind | Silent |
| Are thresholds published for schedule quality? | No pass/fail thresholds; departures are not to be treated automatically as errors (§1.1, §1.3.i) | No numeric thresholds | No numeric thresholds of this kind | Explicitly none — appendix VI |
| Is a schedule risk analysis expected? | Outside the RP's scope | Recommended as part of good practice | Outside the standard's scope | Practice 8: yes, and it sits under credible |
| Is high total float a defect? | Not addressed as a threshold | Not addressed as a threshold | Not addressed as a threshold | Practice 7: unreasonably high total float indicates that logic might be missing or invalid |
| Licence to quote | Member or purchase; may be cited by clause and described, not quoted | Free download; own notice permits extracts with credit | Members only; may be cited by clause and described, not quoted | US Government work; quotable at length |
The divergences, in longer form
Concurrency doctrine
This is the sharpest disagreement of the four and the one that decides cases.
AACE 29R-03 §4.2.D.1 lays out two theories — literal concurrency, which requires the two delays to be in effect at the same time, and functional concurrency, which admits delays that overlap in a looser sense — and endorses neither. It is a methodology document: it describes how each theory is applied and leaves the choice to be argued.
The SCL Protocol's Core Principle 10 does prescribe. It sets a test closer to first-in-time and effective cause than to mere overlap in occurrence, which in practice finds concurrency less often than the functional theory would. The Protocol's own ¶10.10 marks that position as provisional rather than settled — a candour worth noticing, because the position is frequently cited as though it were fixed.
ASCE 67-17 §8.3 goes a third way and states a preference for apportionment where the facts allow it, listing the fact patterns that enable apportionment and the factors to weigh. Where apportionment genuinely cannot be done, §8.2 falls back to a symmetric outcome: excusable, non-compensable, neither party recovering.
Three documents, three doctrines, and the answer for a given project turns on which one the contract adopts. This site's engine takes the governing document as an input rather than assuming one, records the conflict where two documents give different answers, and reports both positions instead of resolving them. It does not decide concurrency.
Concurrency's effect on an extension of time
The same three documents diverge again one step downstream, and not in the same pattern.
- AACE §1.5.B(6) states the gate — the delay must affect the critical path — and is silent on what a concurrency finding then does to the extension.
- SCL CP10 answers directly: the contractor's concurrent delay does not reduce the extension of time otherwise due.
- ASCE §8.3 prefers apportionment, which is a different answer again: the extension is divided rather than granted whole or withheld.
A tool implementing only one of these would be silently wrong on every project governed by another, and its output would look identical either way. The finding here names the gate and the divergence rather than producing a single number.
Pacing notice
AACE §4.2.G records that contemporaneous notices of an intention to pace are very unusual in real projects, and cautions against treating the absence of one as disposing of a pacing claim. SCL ¶15.2 recommends that such a notice be given.
These are not flatly contradictory — one is an observation about evidence, the other a recommendation about conduct — but they produce opposite results when a checker turns either into a requirement. A pacing claim with no notice fails a naive SCL-derived rule and survives an AACE-derived one.
The critical path itself
Everything above assumes the parties agree what the critical path is. They often do not, and the documents do not settle it.
AACE RP 49R-06 gives longest path as its definition, then catalogues four operationally distinct methods in use for identifying the critical path — lowest total float, negative total float, Primavera's Longest Path, and a continuously-valued longest-path variant — and declines to endorse one of them, on the ground that no absolute standard for proper CPM calculation procedure exists. The four are not equivalent: a total-float method can point at a constrained milestone that is not driving completion, and a longest-path method can flag activities a float filter misses.
GAO-16-89G practice 6 takes the longest-duration definition without qualification, which is a coherent choice for an audit instrument and a narrower one than AACE's survey.
Primavera P6 hides the choice behind a single project-level setting, so a P6 schedule reports one answer and discards the other. This project's engine computes and reports both, and names which one a given figure came from. The mechanics are in how the engine computes a schedule.
Where GAO sits
GAO-16-89G is absent from most delay-analysis comparisons because it is not about delay analysis. It answers a different question — is this schedule a reliable forecasting instrument — and it answers two of the questions above that the disputes documents leave open.
On total float, practice 7 states that "Unreasonably high total float on an activity or path indicates that schedule logic might be missing or invalid" — a diagnosis, with no threshold attached. On the critical path, practice 6 makes validity a prerequisite: "Establishing a valid critical path is necessary for examining the effects of any activity's slipping along this path."
And GAO is explicit that it publishes no tripwires. Appendix VI: "No 'pass-or-fail' thresholds or tripwires are associated with the measures. Measures are evaluated in context with qualitative program information and any documented justification." On the mechanical screens that do publish numbers, appendix VII states that "DCMA's 14PA thresholds are not compliance triggers." See GAO-16-89G and the DCMA metrics.
The licence asymmetry, stated plainly
This affects what any honest treatment of these documents can put on a page, and it is why this one reads unevenly.
- GAO-16-89G is a work of the US Government, not subject to copyright protection in the United States. It can be quoted at length, and on this site it is.
- The SCL Delay and Disruption Protocol, 2nd edition is a free, ungated download — scl.org.uk — and its own notice permits extracts where credit is given. Short, credited extracts only.
- AACE Recommended Practices are available to members or by purchase. They can be cited by clause and their methods described. This site never quotes them.
- ASCE 67-17 is available to members. Same treatment: cite the clause, describe the method, do not quote.
The consequence is that the document you can read for free is the one least likely to be quoted faithfully second-hand, and the two documents most often relied on in US practice are the two a reader cannot check without access. That asymmetry is not neutral, and it is worth knowing about before trusting any summary of these standards — including this one. Every clause number above is given so it can be checked against the source.
What this project does with the disagreement
The engine takes the governing document as a stated fact about the project rather than inferring it, applies the one that governs, and records that another document said otherwise. Where the governing document leaves a choice open — the concurrency theory, the criticality school, the analysis interval — the engine reports what was chosen and what that choice implies, and refuses to make the choice itself.
What it computes, what it refuses to decide, and what a reviewing authority must supply are set out in what the result can support, the rule reference and how standards are mapped.
Related: The DCMA 14-point metrics · GAO-16-89G · Comparing delay-analysis methods · Federal construction specifications · State DOT scheduling specifications · Glossary
Source: web/pages/standards-crosswalk.md. Source commit date: 2026-09-11.