A control account is a person.

Here is the mistake, and almost everybody makes it, because it feels like the obvious thing to do.

The cut that has no owner

You have a bill sorted by trade, so you build a budget sorted by trade. Excavation. Piling. Reinforcement, $224,680. Concrete. Formwork. Six neat accounts, and they add to $710,000, and it all reconciles.

Now the reinforcement account overruns. Take that number to a meeting and ask the room a question: whose is it?

Nobody in the building runs reinforcement.

Foreman A runs the substructure — the excavation, the piling, the pile caps, the beams, the raft. He runs the rebar in it, the formwork around it and the concrete poured into it, because that is how a building is actually built. Foreman B runs the frame above ground and does the same.

“Reinforcement” is a column in a bill of quantities. It is not a job. It has a budget and it does not have a human being, and a variance in a box with no human being in it is a variance that nobody fixes.

Same $827,008. Cut it on the trades and you get a list. Cut it on the organisation and you get four people.

That is what a control account is. It is the point where the scope, the budget, the schedule and one name all intersect. Everything else is filing.

Three levels, and they do different work

The confusion that wrecks most systems is that people try to make one level do all three jobs. It cannot.

The control account is where you manage. Four of them on this job. This is where earned value is reported, where the variance conversation happens, and where a person stands up and explains their number. Four is a meeting. Forty is a spreadsheet nobody reads.

The work package is where you measure. Pile cap A. Slab L2 rebar. Sixty of them, each one a single crew doing a single thing, starting and finishing inside a reporting period or two.

The cost code is where you book. Four hundred addresses, and their only job is to catch a timesheet on a Friday afternoon.

Manage at four hundred and you will drown — and a drowning quantity surveyor does not produce careful numbers, he produces plausible ones. Measure at four and you will be blind. Both failures look like diligence from the outside.

Why the work package size is the whole game

Here is the rule, and it is worth more than everything else in this article. A work package must start and finish inside one or two reporting periods.

Not because short is tidy. Because of what happens to measurement when it is not.

A work package that runs for six months has to be assessed as a percentage every month, and a percentage is a judgement, and a judgement made by a man who would rather not have a difficult conversation is worth nothing at all. A work package that starts on the fourth and finishes on the nineteenth does not need a judgement. It is done or it is not. The number is a fact.

We will spend a whole week on this — the six ways of measuring physical progress and which of them are honest. For now, hold the rule: short packages remove opinion from your cost report.

What it would have caught

Go back to the reinforcement disaster. The gang is tying 2.5 tonnes a day against a rate that assumed five, and by month six the labour code is $29,030 over.

Now look at where that number was, before it was a disaster.

Pile cap A is eighteen tonnes. At the estimator's five tonnes a day, the gang has 3.6 days of work and $4,147.20 of labour budget.

They took 7.2 days. They spent $8,294.40.

One hundred percent over, on the first pile cap of the job, on day eight. And notice what else is sitting there: the schedule said 3.6 days and it took 7.2. The cost report and the programme were both screaming the identical thing, in the first fortnight, before a single column had been poured.

At the control account level, that same failure was $29,030 inside a $377,476 account — seven point seven percent, which on a busy job in month two is indistinguishable from rounding. It stayed invisible until month six, by which time 114 tonnes had been fixed at double the labour cost.

Of that $29,030, $24,883 was avoidable — every dollar spent after the first pile cap told you what was wrong.

The work package did not stop the gang from being slow. Nothing in a cost system stops anybody from being slow. It changed how many months you paid for it.

The rules for cutting them

One owner, and he knows he owns it. The test is not whether a name appears in a column of a spreadsheet. It is whether that person, asked in a meeting, says “yes, that's mine” without looking at anyone else first.

Scope, cost and schedule land in the same box. If the budget is cut by trade and the programme is cut by area, you have two systems, and you will spend your career reconciling them by hand. Cut them the same way, once, at the start.

Work packages inside a reporting period or two. Anything longer becomes an opinion.

Work not yet designed goes into a planning package. You will have scope in month eighteen that has no drawings today. Do not pretend to break it into work packages — hold it as one lump with a budget and a rough window, and convert it as the design arrives. An honest lump beats a fictional breakdown.

Practical insight

Take your project's cost report and count the accounts. Then count the people who are actually asked about them in the monthly meeting.

If those two numbers are not close, you have found your problem. Forty accounts and four people means thirty-six of them belong to nobody, and the variances hiding in those thirty-six are the ones that will surface in month eleven.

Then pick the biggest control account and ask its owner one question: what is the first work package that will tell you if this is going wrong?

If he can name it, you are in good hands. If he says “we'll see how it goes”, you have between now and month six.

Key takeaways

✔ A control account is where scope, budget, schedule and one name intersect. No name, no account.
✔ Cutting the budget by trade produces boxes nobody owns. Cut it on the organisation instead.
✔ This job has four control accounts, about sixty work packages and about four hundred cost codes. Each level does a different job.
✔ Manage at the control account. Measure at the work package. Book at the cost code.
✔ A work package must start and finish inside one or two reporting periods, or its progress becomes a judgement.
✔ The slow gang was 100% over on Pile cap A, on day eight, for $4,147 — and 7.7% of a control account, invisible, until month six.
✔ $24,883 of the $29,030 was avoidable. The system did not fail to detect it. Nobody was looking at the right box.
✔ Undesigned scope goes in a planning package. An honest lump beats a fictional breakdown.

What is coming next

You now have four control accounts, sixty work packages, and a budget that closes to $827,008.

All of it is a static number. A pile of money with names on it and no dates.

Next week we spread it across time — and the moment we do, the budget stops being a total and becomes a curve. That curve has a name you already know from Track 1, and this is the week you finally find out where it came from: planned value.