Two decimal places of nothing.

Precision and accuracy are not the same thing, and construction confuses them constantly. A number with two decimal places looks like a fact. It is still only as good as the drawing it was measured from.

What an estimate class actually measures

Cost engineering sorts estimates into classes. At one end, the concept estimate: a floor area, a rate per square metre, half an hour, an answer. At the other end, the tender: a full bill of quantities, every item measured, every rate built.

Each class carries a range. Wide at the start, narrow at the end. Everybody knows this.

What almost nobody internalises is what the class is a statement about. It is not a statement about how much work the estimator did. It is not a statement about the software, the database, or the detail of the build-up.

It is a statement about how much of the scope has actually been defined.

You can spend three months producing a bottom-up estimate to the cent, and if the foundations have not been designed, you have produced a very expensive Class 4 estimate. The drawings decide your class. You do not.

The project you already know

The $1,000,000 project from Track 1 was estimated three times.

At concept it was $880,000 — a floor area and a rate. Everybody in the room treated it as a guess, because it was one, and it carried a range from $616,000 to $1.32 million.

At design development it was $940,000, with quantities off preliminary drawings and a tighter range: $799,000 to $1,175,000.

At tender it was $1,000,000 exactly, off a full bill, with rates built from first principles. Range: $950,000 to $1.1 million. Signed.

It is going to cost $1,267,000.

Read that figure again. The outcome falls outside the tender range. It falls outside the design-development range. The only estimate on that project whose range was honest enough to contain the answer was the one nobody trusted.

This is not an argument for guessing. It is an argument about what happened to the range.

The range narrowed because the drawings got prettier. It did not narrow because the scope got firmer — the bar schedules were still provisional at tender. A Class 1 label was stapled onto a Class 3 body of information, and the estimate stopped telling anyone the truth about how much it did not know.

The weakest input wins

Take the reinforcement package. Two hundred tonnes at $1,446.94. $289,388.

The rate is good. Built from first principles, sourced, checked — call it accurate to three percent.

The quantity came off provisional bar schedules on an undesigned substructure. Call it twenty-five percent, and be grateful.

Multiply them and the package sits somewhere between $210,530 and $372,587. That is the number you are actually carrying. Not $289,388.

Now make a choice. You have some time. Where do you spend it?

Spend three more weeks on the rate. Chase every supplier, re-time every gang, get it from three percent to one. The uncertainty on the package falls by seven percent. You have bought almost nothing.

Or send one detailer to finish the bar schedules and get the quantity from twenty-five percent to five. The uncertainty falls by seventy-one percent.

An estimate is a chain. Chains do not care how strong the good links are.

The costs that are not on the drawing

Now the second half of the problem — and it is the half that swallows contractors.

Add up every measured item in that bill at direct cost and you get $710,000. Every one of them is something you can walk onto site and point at.

The tender was a million.

The difference is not profiteering. It is the site itself. The engineer, the foreman, the offices, the fence, the welfare unit, the tower crane standing over everything, the scaffold, the temporary power, the security guard at the gate at three in the morning. Preliminaries. Nobody drew any of it, and none of it is optional. Twelve percent.

Then escalation. This job runs for two years. Steel, diesel, concrete and wages will not sit still for two years — and last week the steel market moved $140 a tonne while everybody was looking the other way. Four percent, and it is a bet, not a calculation.

Then the company: head office, the bond, the insurance, the bid you lost last month, and profit. Fifteen percent.

Then contingency: the money for the things you know will happen but cannot yet name. Five percent, and we will spend a whole week on it, because it is the most politically abused number in construction.

Two-fifths of that tender is money that is not on any drawing.

And notice the last line. The estimate came out at $998,612 — and somebody rounded it up to a million, because a round number looks better on a covering letter. A rate carried to two decimal places, inside a total adjusted by $1,388 for aesthetics.

Where you put the money

One warning, because it costs people jobs.

Last week's rate was all-in: $1,123.40 of direct cost, with the preliminaries and the margin folded into it, giving $1,446.94. That is one convention.

Most bills of quantities do the opposite. The rate stays at direct cost, and the preliminaries and margin are carried once, in their own sections, for the whole job.

Both give you the same tender — $224,680 of direct rebar, marked up, is exactly the $289,388 you get from the all-in rate. Choose one. Write it down. And never let both happen, because the day they do, you have priced your site overheads twice and you have lost the job to someone who only counted them once.

Practical insight

Find the estimate your budget came from and ask two questions.

First: what class is it? Not what it is called — what it is. Go and look at the drawings it was priced from. If the words “to be confirmed” or “to engineer's detail” appear anywhere in the scope of your biggest package, your baseline is softer than the two decimal places suggest.

Second: where are the preliminaries? In the rates, or in their own section? If nobody can answer immediately, go and check that they are not in both.

You are not being pedantic. You are finding out how much of your baseline is real before the client asks you why the forecast moved.

Key takeaways

✔ Precision is decimal places. Accuracy is scope. They are not related.
✔ An estimate class describes the maturity of the drawings, not the effort of the estimator.
✔ The $1M tender range was $950k–$1.1M. The job will cost $1,267,000 — outside it.
✔ Only the concept estimate, at $880k ±, had a range wide enough to contain the truth.
✔ The estimate inherits the accuracy of its worst input: rate ±3% × quantity ±25% = package ±28%.
✔ Three weeks perfecting the rate buys 7% of certainty. One week finishing the bar schedule buys 71%.
✔ $710,000 of measured work becomes a $1,000,000 tender. Two-fifths is not on any drawing.
✔ Preliminaries live in the rates or in their own section. Never both.

What is coming next

You now have a bill: hundreds of items, each with a quantity and a rate.

You also have a schedule: hundreds of activities, each with a duration and a logic link.

They were built by different people, in different offices, in different weeks, and they do not line up. The bill has an item for “reinforcement, 200 tonnes”. The schedule has fourteen activities that each place some of it.

Next week: how you marry the two — and what happens to your cost report when you do not.