Nobody knows what you have spent.

So let us start where every cost system starts. With one purchase.

One order, five dates

Your project needs reinforcement steel. The buyer places an order: 210 tonnes, $940 a tonne, $197,400.

Day 0, the purchase order is signed. Day 20, the steel is loaded onto a ship. Day 40, it is stacked in the laydown area. Day 45, the supplier's invoice lands in the accounts inbox. Payment terms are 60 days, so on day 105 the money finally leaves the bank.

It is day 50. Your monthly report is due tomorrow.

What is the actual cost of that steel?

Ask four people and you will get four numbers. All four of them are defensible. All four of them are used in real companies, on real projects, this month.

The bank says zero. Not a cent has moved. On a pure cash basis, that steel has cost you nothing at all. It will cost you something in 55 days.

Procurement says $197,400. The order is signed. The company is legally committed to every tonne of it. You could cancel the project tomorrow and you would still be arguing about that money.

Accounts says $118,440. Only 126 tonnes actually arrived. That is the delivery note, that is the invoice, that is the liability sitting in the ledger. The rest is on a second ship.

The site says $59,220. Half of what arrived is in the ground. The other half is a very expensive pile of steel in the laydown yard, and a pile of steel is not a structure.

Now choose the wrong one

Take any of the first three, drop it into your cost report as AC, divide earned value by it, and watch what your CPI becomes.

On the cash basis, your CPI measures your payment terms. Negotiate 90 days instead of 60 and your project looks more efficient. It is not. You have simply moved a date.

On the invoice basis, your CPI measures your supplier's administration department. A slow invoicer makes you look good for a month and then destroys you in the next one.

On the commitment basis, your CPI measures your procurement calendar. Order everything in month two and you will show a catastrophic overrun on a project that has not yet done anything wrong.

Only the fourth number is measuring the project. The cost of the work that has actually been done — matched to the work you gave yourself credit for.

This is the whole of cost control in one line. Everything else in this track is machinery built to obey it.

What it did to a real project

You already know this project. It is the one from Track 1 — a budget of $1,000,000, and by month six a CPI of 0.79 and an EAC of $1.27 million.

Here is what nobody showed you: the month-five report.

At month five, earned value was $250,000 and actual cost was $230,000. CPI 1.09. The project was under budget. That report went to the client, went to the board, and nobody had a single uncomfortable question.

One month later, earned value had crawled to $300,000 and actual cost had jumped to $380,000. CPI 0.79.

Look at the month on its own. $50,000 of work earned, $150,000 of cost booked. The cost performance of that single period is 0.33. Three dollars in, one dollar of building out.

Did the team suddenly become three times worse at pouring concrete in month six? No. $118,440 of that $150,000 was one steel invoice. Almost eighty percent of a catastrophic month was a piece of paper arriving.

The overrun did not happen in month six. It happened in month three, in an air-conditioned office, when somebody signed the purchase order.

Where the money actually went

The bar schedules showed roughly 191 tonnes of steel. The estimate added five percent for waste and priced it at $800 a tonne: 200 tonnes, $160,000. The order that got signed was 210 tonnes at $940. $197,400.

That gap is $37,400, and it is two different failures wearing one coat.

$29,400 of it is price: the market moved $140 a tonne between the tender and the order, and nobody escalated the estimate. $8,000 of it is quantity: the waste allowance went from five percent to ten because the detailing was late and the buyer would rather over-order than stop a gang.

Two decisions. Two different owners. Two completely different conversations — one with a commercial manager, one with a design lead.

Your cost report gave you one number, three months late, with both failures fused into it and neither of them addressable.

That is why cost control fails. Not because the arithmetic is hard.

The three reasons it fails

It arrives late. The ledger is a rear-view mirror with a delay built into it. By the time an overrun is a number, it is history, and history cannot be managed.

It arrives on the wrong basis. Cost measured one way, progress measured another, and a ratio computed between them that means nothing at all.

Nobody owns the difference. The planner owns the schedule, the buyer owns the order, the accountant owns the ledger, the engineer owns the drawing — and the $37,400 belongs to no one, so it belongs to the project.

None of that is fixed by better software. It is fixed by deciding, in advance and in writing, what a cost is, when it exists, and who is standing next to it when it appears.

Practical insight

Open your last cost report. Find the single largest line in this month's actual cost.

Now ask one question about it: when was that money committed? Not invoiced. Not paid. Committed — the day somebody signed something that made it unavoidable.

The gap between that date and today is your blind zone. On most construction projects it is between sixty and a hundred and twenty days.

Everything you decide inside that window, you are deciding on numbers that no longer describe your project.

Key takeaways

✔ “Just ask accounts” is the most expensive sentence in project controls.
✔ One purchase produces four different actual costs: cash, commitment, invoice, and work done.
✔ Only cost of work done can be divided by earned value — the rest measure paperwork.
✔ A CPI of 1.09 became 0.79 in one month because an invoice arrived, not because performance changed.
✔ The single-period CPI was 0.33 — and almost eighty percent of it was one steel invoice.
✔ A $37,400 overrun is two failures: $29,400 of price, $8,000 of quantity. They have different owners.
✔ Cost is incurred when it is committed. It becomes visible months later. That gap is where projects are lost.

What is coming next

We just spent an article on a number: $940 a tonne. And on the estimator's number behind it: $800.

But neither of those is what you actually sold the client. The rate in the bill is for steel fixed in place — and it has a gang in it, a crane in it, a percentage of waste in it, and somebody's guess about how many tonnes a fixer can tie in a day.

Next week we build that rate from nothing, line by line — and find out which of its numbers is actually a price, and which six are opinions.