The site is not lying to you.

He estimated. He stood at the end of a floor, looked down it, and said about two thirds. There is no other way for him to answer the question in the time he has. Nobody is going to tape-measure a painted wall at the end of a shift.

That is the whole subject of this week. The rawest input in project controls is not falsified. It is unmeasured — and unmeasured numbers drift in one direction.

Three ways the same thing happens

Paint is the clearest case because the corroboration is so simple. Material out of the store against area claimed. The two should agree within a sensible loss allowance, and when they don't, one of them is a guess.

Earthworks is the same problem wearing a different coat. Volumes get reported by counting trucks. Twenty loads, so many cubic metres a load, done. But a load is not a fixed quantity, the trucks are not always full, and the conversion factor came from somebody's memory. The weighbridge tickets exist and they say something else.

Manhours are the version nobody sees, because the total is always right. HR has the attendance. What HR doesn't have is where the man stood. Timesheet allocation puts eight hours against an activity because that is what the foreman ticked, and a man who spent half his day helping another gang has just moved four hours of cost onto the wrong work. The total is correct. Every productivity figure built from it is not — and Cost & Cash Week 15 builds earned hours directly on that allocation.

How that number is then turned into a percentage — units, milestones, opinion, cost ratio — is Cost & Cash Week 11, and this week doesn't repeat it. The question here is one step earlier: whether the quantity going into the method was counted or guessed.

Three trades, three mechanisms, one shape. Somebody has to produce a number for work that is too spread out, too fast, or too mixed to measure by the time the report closes.

Why the drift only goes one way

Estimates under pressure are not random. They are optimistic, and consistently so.

Part of that is human — a man reporting his own gang's output is not a neutral observer. But most of it is structural. Nobody is punished for reporting eighty and delivering seventy-five. Everybody is asked why the number went backwards. So the estimate leans forward, week after week, and the gap between what the report says and what is on the wall grows quietly until something forces a reconciliation.

That something is usually the month end. Or worse, the handover, when the last ten percent turns out to be thirty.

The five records that can contradict a claim

You can't re-measure the site. What you can do is find the records that were created for another purpose and ask whether they agree.

Delivery notes and weighbridge tickets. Material arrived, and someone signed for it. This is the strongest corroboration you have because it was produced by a third party with no interest in your progress figure.

Store issues. Material left the warehouse and went somewhere. Paint, cable, welding consumables, formwork — consumption is a proxy for installed quantity, and a good one.

Quantity survey measurements. Somebody measured for payment, on the basis Cost & Cash Week 4 set out. It will be behind the site's claim and it will be more careful, because money depends on it.

Attendance records. How many people were actually on site, which caps how much work can plausibly have been done.

Inspection and test records. Work that has been signed off as complete by someone who had to look at it.

None of these measures progress. Every one of them constrains it. That is the difference, and it is the reason this list is worth more than a better reporting form.

What the planner is actually for

Here is the shift, and it is the argument this entire track is built on.

If you treat the site as your data source, your job is collection: chase the numbers, format them, publish. If you treat the site as one source among several that don't agree, your job changes completely. It becomes reconciliation — finding the number closest to the truth among records that contradict each other.

The second job is harder, slower, and the only one worth having. A planner who only collects is a formatting service. A planner who reconciles is the reason the monthly report can be defended in a room.

One rule makes it work: the planner doesn't correct the number alone. When the store says one thing and the site says another, that goes back to the engineer who reported it and to the department that holds the contradicting record, in the same email, and it stays open until they agree. Not because of politics. Because a figure the site didn't agree to is a figure the site won't defend when it matters.

Put it in writing before you pick up the phone. Write down which number disagrees with which record, copy the departments that hold both, and then ring the engineer and say you have sent something and ask to go through it together. The call is for solving it. The email is the record of how it was solved — and it works the same way when the mistake turns out to be yours. Two days later, two people remember the same call differently. Nobody misremembers a list they both read.

The deadline decides the quality

One thing sets how much of your data is measured and how much is estimated, and it is not the form. It is the hour the report is due.

Ask for yesterday's report by seven the next morning and you have asked the site to report work that finished after they went home. They will estimate, because the alternative is a blank. Ask for it by the afternoon and the same foreman can walk the floor, check with the surveyor, and give you something he counted.

Both projects have a daily report. They don't have the same data in it.

System design

Corroboration is a checking routine, not a form. For each claimed quantity, name the record that was produced by somebody with no interest in your progress figure.

Record	Produced by	Required quality	Verified against	Feeds
Claimed quantity	Site engineer	Measured, or flagged as estimated	Delivery notes, store issues	Progress
Measured quantity	Quantity surveyor	Independent of the progress claim	Site measurement	Valuation, corroboration
Attendance	HR	By area and day	Gate record	Plausibility of the claim
Inspection record	QA/QC	Linked to the activity claimed	Inspection request	Completion evidence
Rules of credit	Project controls	Agreed before work starts, not after	Baseline	How a quantity becomes a percentage

None of these measures progress. Each of them constrains it, and a claim that survives all five is one you can defend without re-measuring anything.

Practical insight

Take your own last month's progress and pick the three activities with the largest reported quantities. For each, ask one question: what record, produced by somebody outside your reporting chain, would agree with this number?

Sometimes you have it immediately — a weighbridge ticket, a delivery note, a signed inspection. That activity is measured, and you can defend it in a room.

Sometimes there is no such record at all. That one is an opinion with a percentage sign on it, and you have just found where your curve is going to correct itself later.

Then do it once more, deliberately, on the activity closest to your critical path. That is where an optimistic estimate does you the most damage, and it is almost always the one nobody has corroborated.

Key takeaways

✔ Site data is rarely falsified. It is unmeasured — estimated because the area is too large or the deadline too early to count.
✔ Unmeasured numbers drift one way. Nobody is questioned for over-reporting; everybody is questioned when progress goes backwards.
✔ Paint against store issues, volumes against weighbridge tickets, hours against where the man actually stood — three trades, one mechanism.
✔ Manhour totals can be right while every productivity figure built on them is wrong, because allocation is what breaks.
✔ Five records can contradict a claim: deliveries, store issues, QS measurement, attendance, inspection sign-off.
✔ None of them measures progress. Each of them constrains it, and constraint is enough.
✔ The planner reconciles rather than collects — and never corrects a number alone. It goes back to the reporter and the record holder together.
✔ The report deadline decides how much of your data was counted and how much was guessed.

Records born here. Progress claim sheet · corroboration note · the query email when two records disagree.

What is coming next

This week was about a number that already exists, and whether you can believe it.

Next comes the number that does not exist yet: the work the site says it is ready to start. Ready means six different things to six different people, and the gap between them is where a look-ahead stops being a plan and becomes a wish.

Next week: workfront and the six readiness tests — drawing, material, access, permit, labour, plant.