First published on LinkedIn by Morne Beeslaar. Read it on LinkedIn. This version carries corrections to the clause references and claim-file facts made after first publication.
Open FIDIC Red Book 1999 at Sub-Clause 20.1. The Contractor must give notice to the Engineer as soon as practicable, and no later than 28 days after becoming aware, or after the point at which it should have become aware, of the event. Miss that, and three things happen at once. The Time for Completion is not extended. There is no entitlement to additional payment. The Employer is discharged from all liability in connection with the claim.
Read the third one again. Discharged from all liability. Not reduced. Not disputed. Discharged. And note that the 28 days is a longstop, not a standard: a contractor who sits on a known event for 26 days has already failed the primary obligation.
FIDIC 2017 restructured Clause 20 and put the same trap at Sub-Clause 20.2.1, with two changes worth knowing. The bar now runs against whichever Party is claiming, so the Employer is caught by its own clock. And the Engineer must challenge a late Notice within 14 days, with reasons, or the Notice is deemed valid.
The Notice of Dissatisfaction is not new either, whatever you may have been told. FIDIC 1999 already required one within 28 days of the DAB's decision before either Party could arbitrate, and 2017 keeps that at Sub-Clause 21.4.4. What 2017 added is a second one, earlier, at Sub-Clause 3.7.5, against the Engineer's determination, on the same 28 day fuse. Say nothing for 28 days and the determination is final and binding on you.
So the later edition is not simply harder or softer. It loosens the first fuse: a late Notice can survive if the Engineer stays silent for 14 days, and even a challenged one can be treated as valid in the determination, weighing prejudice to the other side and what it already knew. Then it adds fuses further along. A claim whose contractual basis is not set out within 84 days lapses. A determination nobody objects to within 28 days binds. The numbers move between forms and editions. The mechanic, a clock on every step, does not.
The failure mode comes before the product
On a live contract, nobody misses a notice deliberately. They miss it because the event that triggered the clock did not look like a claim on the day it happened.
Late access to a work face. A drawing revision that arrives with no covering instruction. A site instruction given verbally in a Tuesday meeting. Each one is ordinary. Each one starts a clock that nobody started counting. Four months later the delay is obvious, the cost is real, the entitlement was there, and the notice window closed on a day that nobody can now identify.
The claim then fails on a date, not on a fact. That is the expensive way to learn what your contract says.
A notice is not an entitlement
The time bar is the first way to lose a claim. It is not the only one. Here is one, anonymised: a claim file from a process-plant package on a mining expansion, where we sat on the owner's team.
Five months after award, the contractor served its first extension of time claim. Owner-supplied electrical rooms were arriving after the contractual completion date, and they sat on the critical path. The contractor's forecast moved 76 days. That claim was well made. The delivery dates came from the owner's own expediting report, and the effect on completion was traced.
Over the next six months three more claims followed, each stacked on the last. 46 days for late owner-supplied steel. 23 days for updated delivery dates. 8 days for added scope. Cumulative slip claimed: 153 days.
The time bar was never the issue on this file. The issue was proof. Two of the four increments were recommended for rejection and a third was held as premature: one because the effect was asserted, not demonstrated, one because it counted an event already inside the first claim, and one because the scope it depended on had not yet been defined.
The reason was the programme. The quality tool reported every activity in it as critical, which means it could not show what any single event actually drove. It carried open ends, 98 negative lags and a heavy load of constraints, and it failed a standard schedule-quality check at every submission. The Baseline Execution Index, which compares the tasks actually finished with the tasks the baseline said should be finished by then, fell from 65 percent at the second submission to 12 percent at the third. The accepted indicator of a programme that can hold its dates is 95.
A notice preserves the right to claim. It does not prove the claim. The programme proves it, or it does not.
The owner's side of the same file
The owner was not clean either, and this is the half most vendors leave out.
On its third submission the contractor put in writing that it had never received a formal approval or rejection of the revised programme it had sent two months earlier. The delays driving the critical path traced to items the owner itself was supplying. When the claim settled, the new completion date landed on the contractor's third submission, the one whose added days our review had recommended rejecting.
The money followed the same shape. The settlement came in at just over half of the amount claimed, and just under the figure the owner's own entitlement analysis had put on it.
Then the part that should worry both sides. The owner's settlement position was that every impact known or foreseeable at the date of the claim was absorbed into the new completion date. Anything the contractor had not registered by then was gone. Anything the owner had not priced by then was in the date for good.
Nobody lost that claim on day 29. The contractor's number was halved on proof. The owner gave away time on silence and on its own late supply. Neither is a legal problem. Both are register problems.
What we built
Forseti is the contracts and claims intelligence agent in the FaolanIQ fleet. It does five things.
- Reads the executed contract and produces a risk register mapped to specific clause numbers, not to themes
- Builds a notice and time-bar schedule for that contract, so every obligation with a clock on it is visible before the clock starts
- Flags the deviations from standard form: amended liability caps, fitness-for-purpose obligations imported into a Red Book, risk accepted in the particular conditions that the general conditions gave away
- Structures a claims narrative in four parts, entitlement, causation, effect, quantum, each anchored to the clause it relies on
- Searches the contracts corpus for the precedent and the comparable position
The output that clients look at first is the plainest one. A one-page map of every notice period in their own contract, with the clause number and the edition next to it.
The file above is the reason the register does not stop at the notice. Each entry carries the event, the clause, the date the clock started, the longstop, whether notice went, and whether the effect has been demonstrated against a programme that can actually demonstrate it. Four claims, one register, and the gap between "claim served" and "effect shown" is visible on the day it opens, not on the day of settlement.
Why clause numbers matter more than summaries
A contract summary is a comfortable document. It reads well and it defends nothing.
The test is whether a proposition can be traced back to the words that carry it. If the register says a claim is barred 28 days after the contractor knew or should have known, it says FIDIC 1999 Sub-Clause 20.1 next to it, and the reader can open the book and check us. Edition included, because a register built on 1999 numbering and applied to a 2017 contract will point at the wrong clause and miss that the bar now runs both ways.
If the risk register scores a clause as onerous, it shows the standard-form position and the amended position side by side, so the reader can see what was traded and decide whether they meant to trade it.
That is the same argument we make about every product in the fleet. The finding is not the deliverable. The traceable derivation of the finding is the deliverable, because that is the part that survives contact with the other side.
It cuts the other way too. When we checked this article against the standard forms before publishing it, the check caught our own drafting: an earlier version had 2017 introducing a Notice of Dissatisfaction path. It did not introduce it. A second pass then caught the check itself, which had missed that 2017 lets a late notice be rescued in the determination. A register that cannot catch that in its own author is not a register.
What it does not do
This is contract intelligence. It is not legal opinion, and Faolan Consulting is not a law firm.
The line is not a disclaimer bolted to the bottom of a report. It is a scope boundary that changes what gets produced. Forseti will tell you that Sub-Clause 20.1 bars a claim after 28 days and show you exactly where that sits in your document. It will not tell you whether that bar is enforceable against you in a South African forum on your facts. That question goes to instructed attorneys, and the work product is built so that it hands over cleanly when it does.
It also does not replace the contracts professional or the planner. Reading a contract is not the hard part of contract administration, and reading a programme is not the hard part of a delay claim. Keeping the register and the programme honest across months of live events is the hard part, and that is the part that goes wrong first when a team gets busy.
The number that matters
Twenty eight days is the one on the page. It is knowable in advance, on day one, for every clause in the contract you have already signed.
Twelve percent is the one in the file. It is the number that told the owner the programme behind the last three claims could not prove them, months before anyone sat down to settle.
If you cannot produce your own notice schedule from your own contract this week, and show which of your open events have a demonstrated effect behind them, that is the finding.
Contract intelligence only, not legal opinion. Faolan Consulting is not a law firm. The claim file described is anonymised; no party, project or jurisdiction is identified.
Faolan Consulting. AI-Integrated Project Controls and Planning.
Controls. Intelligence. Delivered.