The last 5% takes a year.

This is the most predictable failure in construction, and almost nobody plans for it.

Two gates, not one

Start with the mistake underneath all of it.

Every task in the schedule is marked 100%. Every bar is green. The amateur conclusion — and it is an amateur conclusion, however senior the person making it — is that the project is therefore finished.

There are two separate gates, and they have very little to do with each other.

The inside gate is a technical audit: is every item in the work breakdown structure actually built, and did it pass quality? Not marked complete — built, and inspected.

And a warning here that traces straight back to Week 21: you audit against the WBS as it stands today, with every approved change order integrated. Not the original scope statement. Check the finished building against a three-year-old document and you'll find "extras" that were properly approved and "gaps" that were properly deleted — and you'll spend a fortnight arguing about ghosts.

The outside gate is a signature. The client has formally accepted the deliverables. That's it. That's the whole test.

Pass the first and assume the second, and you release your team into a snagging list you haven't seen yet. Every item the client raises is now unpaid rework — done badly, by whoever is left, because the people who built it are on another site and are not coming back.

The software says done. Only the client gets to say finished.

"Done" has to be decided in advance

Which raises the question that determines whether the outside gate is a formality or a nightmare: what exactly does the client have to see before they sign?

If that's being worked out in month thirty-four, you have already lost.

Here's how it goes wrong, and it's never dramatic. Eighteen months ago, in a routine meeting, someone said "yes, of course we can sort that out for you." It was friendly. It was meant well. Nobody wrote it down.

The client remembered. They have been assuming it for a year and a half. And on handover day they will not sign, because the thing they were promised isn't there — and from where they are standing, that is entirely reasonable.

Acceptance criteria belong in the contract, agreed and signed, before anyone breaks ground. "Done" has to be a fixed target. The moment it becomes a matter of opinion, it becomes a matter of whoever is most determined — and at handover, with retention money on the table, that will not be you.

Nobody is left to close it

Now the structural problem. It isn't anybody's fault; it's just how organisations behave.

The moment the last major milestone lands, your best people become available — and the business notices instantly. They get pulled onto the next job, because that's where the revenue is, and closeout looks like tidying up.

So the administrative workload spikes at the exact moment your capability collapses. Final accounts, as-built records, sign-offs, the archive — all landing on whoever was slowest to escape.

There's a quieter version of the same problem on site. On a long project, the team has built a life around it: the routine is comfortable, the people are familiar, and nobody is entirely certain what they're doing next. So the last few work packages develop a strange, unspoken drag. Nothing is refused. Things just take a little longer than they should.

Both of these are predictable, which means both can be planned for — and the fix is almost embarrassingly simple. Put the closeout in the schedule. Named activities, real durations, real people, resourced and protected like any other work.

If closeout isn't in the plan, it isn't work. And if it isn't work, it will be done by nobody, badly, in November.

The data you throw away

And finally, the part that costs your company the most and is noticed by absolutely nobody.

Three years of hard evidence about how your organisation actually builds. What things really took. Where the estimate was wrong. Which subcontractor never recovered from a bad start. How much rain you genuinely lost. This is the single most valuable dataset your business owns.

Almost all of it gets thrown away, and here is exactly how.

You hold a lessons-learned meeting. At the end. In a room. Where you ask people to remember three years of decisions — and the person who could actually explain why the steel was late in April left the company in March.

It cannot work. It has never worked. It's an archaeology exercise dressed up as a process.

A lesson has a shelf life, and it's about a fortnight. Written the week it happened, by the person it happened to, it is precise, honest and useful. Recalled eighteen months later in a conference room, it is a vague sentence that helps nobody.

So capture it continuously. When a duration turns out badly wrong, write down why — that month, in the schedule, while you still know. When an approach works, say so. It costs ten minutes, and it is the only mechanism by which your next estimate becomes better than this one.

Remember Week 12: the best duration estimates come from historical data. This is where historical data comes from. A company that closes out properly is a company whose next bid is sharper than its competitors' — not because it is cleverer, but because it remembers.

A project ends three times: when the work is built, when the client signs, and when the knowledge is captured. Most organisations manage the first, argue their way through the second, and quietly skip the third — then wonder why every project feels like the first one.

Practical insight

Open your schedule and look for the closeout activities.

Not "handover" as a single milestone with a diamond on it. Actual activities: final account settlement, as-built drawings, client acceptance, archiving, lessons captured. With durations, with resources, with names.

If they aren't there, they are not going to happen — and the person who ends up doing them, alone, in November, may well be you.

Key takeaways

✔ Ticking every task to 100% is not completion — it's one of two gates.
✔ Inside: is it built, did it pass QA. Outside: has the client signed.
✔ Audit against the current WBS with all approved changes, not the original scope statement.
✔ Release the team after the internal gate and every snag becomes unpaid rework.
✔ Acceptance criteria go in the contract before work starts. "Done" must be a fixed target.
✔ Closeout workload peaks exactly as your best people are pulled onto the next job.
✔ Put closeout in the schedule with names and durations, or it gets done by nobody.
✔ A lesson has a shelf life of about a fortnight — capture it during, never at the end.
✔ Your project history is where next year's estimates come from. Most companies bin it.

What's coming next

The project is closed, the knowledge is banked, and every number we've built over twenty-five weeks is honest.

And none of it matters if the person who has to act on it doesn't understand what you sent them.

Next week: reporting, and the altitude problem. The site manager needs next week in detail. The board needs the whole project in one line. Send either one the other's report and you have achieved nothing — except a reputation for producing documents nobody reads.