Strip away the terminology and every delay determination — by an engineer, an adjudicator, a dispute board or a tribunal — reduces to three tests. Understanding them changes how you build the file, and most of what follows in this field guide is really about those three tests wearing different clothes.

1. The three questions behind every determination

Strip away the terminology and every delay determination reduces to three tests. The order matters. I have watched contractors invest six figures in a forensic analysis for a claim that failed the first test on a notice provision they had never read carefully. Spend the first day on the contract, not the schedule.

Table 0 — The three tests behind every delay determination. Entitlement in principle precedes causation, which precedes quantum.

2. The monthly discipline that wins claims before they exist

If you take one thing from this article, take this list. It costs a planner perhaps four hours a month and it is the difference between a claim that assembles itself and one that has to be excavated years later at ten times the cost.

Table 1 — The monthly delay-readiness routine. Four hours a month is the difference between a claim that assembles itself and one that has to be excavated.

⚠ A lesson learned the hard way. On a wastewater upgrade in 2019, the team ran a single live programme and overwrote it each month. When the claim came, we had one baseline, one final schedule and twenty-two months of nothing in between. Reconstructing the intermediate positions from progress reports and invoices took eleven weeks and produced a result that the employer's expert could legitimately challenge on three separate grounds. The archive would have taken about ninety seconds a month.

3. Modelling delay into a live programme

When you are seeking relief during execution rather than after completion, the analysis is prospective: you take the accepted programme updated to the moment before the event, insert the event as a sub-network — a fragnet — and re-run the schedule to see where completion lands. The chart below shows three fragnets modelled into the hospital programme.

Figure 1 — Fragnets modelled into the live programme, Riverside Hospital (illustrative)

Figure 1 — Delay fragnets inserted into the accepted programme at their respective impact dates. Contract completion sits at day 440; the impacted position lands at day 511.

Three things make a fragnet defensible. It must be logically tied to the specific activities the event actually touched, not floated in as a bar with a start-to-start link to the project. Its duration must come from a document — the number of days a response took, the shipping manifest, the revised drawing issue date — and not from an estimate. And it must be inserted at the impact date, meaning the date the event first affected the works, which is frequently later than the date the event occurred.

That last point catches people out. A design query raised on 3 March that was not needed on site until 19 April has an impact date of 19 April. Inserting it on 3 March overstates the delay and invites a reviewer to recalculate the whole analysis themselves.

4. Two families of method, and where they belong

Rather than memorise six technique names, it helps to see them as two families that answer slightly different questions. Time impact analysis sits in the prospective family; windows analysis and its relatives sit in the retrospective one. The full taxonomy is covered in the encyclopedia entry on forensic delay analysis.

Table 2 — Prospective versus retrospective method families. In practice, both often appear in the same file.

The contractor claims prospectively during the works under the notice regime, and the final position is reconciled retrospectively at the end. Where the two produce different answers — and they will — the retrospective result generally carries more weight after completion, because it is observing rather than predicting.

5. Evidence is a hierarchy, not a pile

A recurring mistake is treating a large appendix as strong evidence. It is not. Reviewers weight evidence, and knowing the weighting tells you where to spend your effort.

Table 3 — How delay evidence is weighted in practice. Spend your effort at the top of the table, not on the thickness of the appendix.

Expert tip ✔ Build the event register before you build the analysis. One row per event: reference, description, source document and date, activity IDs affected, impact date, end date, party at risk, notice reference. On the hospital project that register was 94 rows and it became the index for the entire submission. Every table and every fragnet in the report pointed back to a row number, so the reviewer never had to search for anything.

6. Classifying critical delay: time, money, both or neither

Once you have measured critical delay in each window, every day has to be placed in one of three buckets. This is where a submission either demonstrates discipline or reveals that it is simply asking for everything.

Table 4 — The three buckets. Every measured critical day lands in exactly one of them, with time and cost entitlement decided separately.

Figure 2 — Critical delay classified window by window

Figure 2 — Critical delay days classified window by window. April is the interesting month: nine contractor days alongside seven excusable non-compensable days.

Look at April in that chart. Nine non-excusable days sit alongside seven excusable, non-compensable days. That is a concurrency picture, and the honest reading is that the contractor is protected from liquidated damages for the overlap but recovers no prolongation cost for it. Presenting April as a clean seven-day compensable claim would be the fastest way to have the entire submission read sceptically.

Across the seven windows the totals came out at 45 excusable days, of which 34 were also compensable, against 23 days of contractor-risk delay. When we tabled that split ourselves in the opening summary, the conversation with the employer's team changed character immediately. We were no longer arguing about whether the contractor accepted responsibility for anything; we were arguing about eleven specific days in two windows.

7. Float erosion: the earliest honest warning

Total float is the most underused early warning indicator in project controls. A path does not become critical on the day it goes negative; it announces its intention over several months as float drains away. If you plot float by path across successive updates, you can see the problem forming — the same mechanics behind float management and the negative-float failure mode we dissected in the Primavera P6 negative float case study.

Figure 3 — Float erosion is the earliest honest warning you get

Figure 3 — Total float on two paths across twelve updates. The facade and MEP path signalled its trajectory long before it turned negative at update 8.

On the chart above, the facade and MEP path was losing three to seven days of float every update from update 3 onwards. By update 6 the trend was unmistakable. Notice was actually issued at update 8, when the path went negative — two months later than the data justified. Those two months of silence became the employer's strongest argument for reducing the award, and they were entirely avoidable.

Float ownership deserves a word here too. Many standard forms operate on the basis that float belongs to the project rather than to either party, so relief is granted only where the delay consumes all available float and pushes completion. Some bespoke amendments say otherwise. Read the clause, then state your position on float explicitly in the methodology — silence on the point is read as an assumption in your own favour.

8. Notice regimes: the trap that has nothing to do with scheduling

Notice provisions defeat more meritorious claims than bad analysis does. The mechanics differ substantially between standard forms, and the differences are worth knowing before you need them.

Table 5 — Notice and assessment mechanics across common standard forms. Orientation only; the executed contract, including amendments, always governs.

A practical habit: on mobilisation, produce a one-page notice map for the project team listing every notice obligation, the trigger, the period, the addressee and the required form. Pin it in the site office. It takes an afternoon and it has saved claims worth millions on projects I have worked on.

9. Common mistakes

Table 6 — The mistakes that sink otherwise sound submissions, and the correction for each.

10. Expert tips from live submissions

• ✔ Open with a one-page summary that states the method, the total slip, the excusable days, the compensable days and the contractor-risk days. Reviewers read that page and skim the rest. If it is honest, the rest is read charitably.
• ✔ Keep the working-day and calendar-day conversion in a single visible table. Ambiguity here is the most common source of pointless argument in the review cycle.
• ✔ Photograph and date-stamp obstructed work faces on the day the obstruction occurs. A dated image of an area that could not be accessed is worth more than three pages of narrative.
• ✔ Where records are missing, name the gap in the methodology section and explain how you bridged it. Every reviewer finds gaps; only some of them find them before you disclose them.
• ✔ Have a colleague who has never seen the project attempt to follow your analysis from the event register to the final number. If they cannot, neither can the engineer.
• ✔ Consider automating the mechanical layer — file parsing, window comparison, critical path change detection, slip quantification. The time you save goes into causation, which is where claims are actually decided.

Key takeaways

• 📌 Claims are won during the project, through archived updates, dated event logs and timely notices.
• 📌 Test the contract first. Entitlement in principle precedes causation, which precedes quantum.
• 📌 Prospective methods answer 'what will this do'; retrospective methods answer 'what did happen'.
• 📌 Evidence is weighted. Contemporaneous native records outrank everything created after the dispute.
• 📌 Classify every critical day as compensable, excusable-only, or contractor risk — and show your workings.
• 📌 Float erosion warns you months ahead of a negative float alarm. Watch the trend, not the threshold.
• 📌 Concede your own delays early. It costs a few days and buys credibility on the days that matter.

🔗 Related career guides

• 📘 EOT claim preparation steps: building an extension of time that actually gets awarded — the seven-stage claim sequence that this field guide's record discipline feeds into.
• 📘 How to become a delay analyst — the forensic career built on exactly these methods, evidence tiers and notice regimes.
• 📘 How we fixed -19 days of negative float in Primavera P6 — the schedule mechanics behind the float erosion warning signs in Figure 3.
• 📘 Inside the Riverside office building Primavera P6 tutorial — an end-to-end P6 build that rehearses the baseline, update and fragnet hygiene this guide depends on.
• 📘 Schedule entropy: why great schedules slowly become chaotic — the drift that quietly destroys the evidential value of programme updates.

Companion reading in the encyclopedia: Forensic Delay Analysis, Time Impact Analysis, Windows Analysis, Baseline Schedule, Float Management, Critical Path Method and Liquidated Damages.

Final word

The hospital submission that was resolved in three weeks did not contain a single technique that is unavailable to any planner reading this. It contained a dated archive, a disciplined event register, notices issued on time, and an author who was willing to write down the twenty-three days that belonged to their own side. The analytical sophistication was ordinary. The record-keeping was not.

If you are on a live project right now, the highest-value thing you can do this week is not to learn a new delay analysis method. It is to check whether last month's programme update still exists as a separate, dated file — and if it does not, to start the archive today.

Dr. Hassan Khames Eliwa, PhD, works in project controls across building, healthcare, transport and infrastructure programmes, specialising in planning and scheduling, earned value management, forensic delay analysis and the preparation and defence of extension of time claims.