Valeur.A good claim is lost on procedure, not on merit.
A claims director does two jobs, and the two do not live on the same calendar. One is the entitlement itself: the event happened, the loss is real, the figure is defensible. The other is procedure: did the notice go out in time, is the record contemporaneous with the event, which document carries which line. The first makes you right; the second makes being right usable. This page is about what the second one is attached to — the evidence list, the method behind the calculation, and where each line came from, all held in one file.
A sixteen-factor productivity table and an eight-document evidence list; the figures here were read out of the product’s own source while this page was written.
- 16 productivity factors in one named table
- 3 severity steps per factor — 48 coefficients
- The result is written as 5 named money lines
- The automatic pass writes 4 of the 16; you flag the rest
A defensible loss stops being defensible in five places.
These are not stages a claim passes through in order. They are five openings, and all five are in the same file at the same time. Each one closes on the sentence that will be said back to you across the table.
-
01
The clock starts on the day you file, not on the day it happened.
The event happened and the loss started. Your cash flow felt it that same week. But what gets measured as the “length” of a dispute is counted from the moment the claim is formally submitted under the contract. The months in between sit in your ledger and in nobody’s average.
The notice calendar holds the event date as a field of its own. It counts the remaining days from there. The 20.1 and 8.4 records carry the 84-day detailed-claim obligation that follows the notice inside the same record.
RuleThe distance between the event date and the submission date becomes a number you read off the file itself. Shortening that distance is still your job. What the structure here does is keep it in view.
-
02
You can feel the productivity loss; you cannot build its arithmetic.
The crew slowed down, everybody knows it, nobody can attach it to a figure. A loss that cannot be attached does not enter the claim file. What enters is a loss with a name and a severity.
The MCAA cumulative-impact table holds sixteen factors in one named list. Each factor carries three severity steps — minor, average, severe — which is forty-eight coefficients, and in all sixteen the steps rise from minor to severe. The calculation sums only the factors marked as detected.
ScopeThe factor’s name, its severity and that severity’s coefficient stand as their own line in the output, one line per factor. An unrecognised severity falls to the smallest of the three steps. Not the largest. The method’s name and its two legal references — MCAA Bulletin No. PD2 and Clark Concrete v. GSA, VABCA (1998) — are written inside the output itself. Those are the references the output carries, not a legal view of our own.
-
03
The other side says the factors overlap.
The objection to a multi-factor claim is known: double counting. The other side’s expert says it in the first session, and the defence is expected to be yours to build.
The correction runs when factors overlap. When more than one factor is flagged, the calculation applies a fixed-rate overlap correction on top of the gross loss and writes it into the output as its own line item: the rate in one field, and a note saying how many factors overlapped in another. On a single-factor file the correction stays out, and the output reports that in the same place.
ScopeThe correction works on the gross loss; overhead and profit are added afterwards, on top of the corrected figure, and together they come to more than the correction does. So this is not a discount — it is a line item standing in the open, ready to be argued with: the answer to the double-counting objection is inside the file, written before you get there. All three rates are in the table below, side by side.
-
04
The method carries exactly as much weight as your reference period.
All of the measured mile’s weight comes from the consistency of its reference period. Pick a volatile one and the method works against you. The other side is who notices first. The output also carries a label for this method’s evidentiary standing; that label is the output’s own text rather than a legal view of ours.
The AACE 25R-03 calculation measures the reference period’s coefficient of variation and returns it as a field of its own. The result is tied to one of three reliability bands. The lowest band says it outright: choose a different reference period. Where the reference or impacted data has not been entered, the calculation names which side of the comparison it is waiting for instead of producing a figure.
ScopeThe coefficient is a number that gets measured and shown. It is not a seatbelt. A one-day reference period lands in the top band, because the deviation of a single point is zero by construction. Reading the band means reading how many days were measured alongside it — and that is in the same output, in its own field.
-
05
The evidence list gets finished after the file closes.
There is one kind of evidence you cannot produce later. It is the contemporaneous record. Learning which document you needed while you write the claim means the window for collecting it is already behind you.
The claim package arrives with an evidence list of eight core documents under three headings. The list is not fixed: flag the overtime factor and overtime authorisation forms are added as critical; flag season and weather and official meteorological data is added the same way. Flag both and the list runs to ten items.
ScopeTwo of the sixteen factors can grow the list, and those two sit inside the four the automatic pass writes. The only critical item in the base list is the 20.1 notice record. The other seven are marked to collect. The list uses exactly two status values: critical and collect.
Every figure in this section — including the clause numbers, the factor and coefficient counts, the days that follow the notice, the items on the evidence list and the numbers inside the citations — was read by parsing the product’s own source while this page was written. None of them is a customer outcome.
What is published is not what is in your file.
The two findings below come from published reports, and both are averages over files that are already in dispute — both say so in their own text. The box underneath them is our work. The seam between the two is left visible on purpose.
On contracts that went into dispute, the sums in dispute averaged 33.4% of the contract budget.
HKA, CRUX Insight — 8th Annual Report, 2025. More than 2,200 projects, 114 countries. Denominator caveat: this pool is made of projects that experienced a dispute, not of all projects — it cannot be read as “a third of every contract”.
In North America a construction dispute ran 12.5 months on average.
Arcadis, 15th Annual Construction Disputes Report, 2025 (2024 data). Scope: North America. The average is taken over files that went into dispute.
That is what the sector publishes. What follows is the structure working inside your file. Productivity loss is tied to one sixteen-factor table; each factor carries three severity steps, and only flagged factors enter the sum. When more than one factor is flagged, a fixed-rate overlap correction is applied on top of the gross loss and written into the output as its own line; overhead and profit are added after it. On the measured-mile side the reference period’s coefficient of variation is measured and the result is tied to one of three bands. The claim package arrives with an eight-document evidence list that can grow by two items depending on which factors you flag. This is the shape of the list and the calculation — a structure rather than a result.
The two figures above stand with their own years on them. Where the report starts its clock depends on the report’s own definition, and that definition is left off this page: the copy we read word for word comes from the 2015 edition, while the figure above comes from the 2025 one. Search says the same wording recurs from edition to edition — but “search says” and “we read that line in that edition” are not the same sentence. Rather than build one sentence out of two editions, we leave the figure with its own year.
The same rule, met from the other end.
A claims director meets this rule after it has already decided whether the rest of the file is worth building. The same two principles stand here in the same words. Fresh ones would read as a second, independent finding.
In some standard forms these notices are expressed to be conditions precedent to entitlement.
Society of Construction Law, Delay and Disruption Protocol, 2nd Edition, 2017.
A notice given in accordance with the condition precedent keeps the right to claim alive; where the condition is unmet, the right falls away.
Pinsent Masons, Out-Law Analysis, 2023.
The notice calendar carries that principle as a field: three of the twenty-three clauses carry a second flag marking the bar absolute — 8.4, 19.1 and 20.1. The one critical item in the base evidence list is the 20.1 notice record as well; the two lists meet on the same clause. The record’s own term is an absolute time bar. The legal characterisation above belongs to the citations rather than to the record.
We use no percentage in this section either. How many claims fall to a time bar is not something we hold as a published rate; the data stays inside private arbitration. A plausible-looking number placed here would read as though it came from the two citations beside it.
The figure is built row by row.
The numbers that build the figure are in the rows below. The sector rates, the time bars and the citation numbers live in other sections, with their own sources. Rows four, five and six are a set: reading the overlap correction on its own means reading it without the two items added after it.
- Productivity factor table
- 16 factors in one named list
- Severity steps per factor
- 3 steps — minor · average · severe. 48 coefficients; in all sixteen the steps rise from minor to severe
- Factors that enter the sum
- Only those marked as detected; each one gets its own output line with its severity and coefficient
- Overlap correction
- −15% of the gross loss above one factor; on a single-factor file the output reports that in its own note
- Overhead
- 15% of the corrected figure, added after the correction
- Profit
- 8% of the corrected figure, added after the correction
- Money lines in the output
- 5 named lines — gross · net · overhead · profit · total
- Evidence list
- 3 headings, 8 core documents; the one critical item in the base list is the 20.1 notice record, and two conditional additions take the list to 10
Each row in this table was read while this page was written by parsing the product’s own source; the rates were checked on top of that by actually running the calculation over a two-factor and a single-factor file and counting the result.
Five questions a claims director asks.
Which productivity methodology do you use?
MCAA cumulative-impact analysis: one sixteen-factor table, three severity steps per factor, forty-eight coefficients in total. The method’s name and its two references are written inside the output — MCAA Bulletin No. PD2 and Clark Concrete v. GSA, VABCA (1998). Those are the references the output carries, not a legal view of our own. The AACE 25R-03 measured mile stands alongside it as a separate calculation.
What do you say to the double-counting objection?
The correction is a step inside the calculation. When more than one factor is flagged, the calculation applies a fixed-rate overlap correction on top of the gross loss and writes it into the output as its own line item: the rate in one field, a note saying how many factors overlapped in another. On a single-factor file the correction stays out and the output reports that too. To see exactly where the correction moves the total, read rows four, five and six of the table above together: overhead and profit are added after the correction, on top of the corrected figure.
How will I know whether my reference period suits a measured mile?
The reference period’s coefficient of variation is measured, returned as a field of its own, and tied to one of three bands; the lowest band tells you outright to choose a different reference period. The band should not be read on its own: the coefficient depends on how many days were measured, and a one-day reference lands in the top band because the deviation of a single point is zero by construction. The number of days measured is in the same output.
Does an AI write the legal argument?
With a language model connected, a draft argument is produced and the output says so in a field of its own. Without one, the output returns a fixed three-item basis in place of the draft and reports the situation in that same field: the 20.1 notice obligation, the reference for the MCAA methodology, and contemporaneous records as the evidentiary base. The quantum calculation is identical either way — deterministic. What depends on the model is the narrative alone.
Who tells me which documents to collect?
The evidence list arrives with eight core documents under three headings and grows with the factors you flag: the overtime factor adds authorisation forms, the season-and-weather factor adds official meteorological data, both as critical. Two of the sixteen factors can grow the list; flag both and it runs to ten items. The one critical item in the base list is the 20.1 notice record. The list uses exactly two status values: critical and collect.
Let’s reopen a file you have already closed.
Thirty minutes. Bring a claim that has been settled; we will read together which of the sixteen factors would be flagged, which line the overlap correction shows up on, and how the evidence list would have grown on that file.