Cost is committed at the moment somebody decides. It is reported at the moment somebody invoices. Control lives in the gap.
Last updated
Zepth Core module
+80%
large construction projects running over budget
McKinsey Global Institute — Reinventing Construction
Up to 80% over budget, and around 20% longer than scheduled. Both are shares, not durations. The overrun is not one bad decision; it is the accumulation of decisions nobody costed at the moment they were made.
Unawarded balance
current budget − committed − approved variations − direct costs. The one number that answers “how much is still mine to control?”
Zepth cost-ledger model
A doctrine, not a statistic. It is the figure the classic budget-versus-invoices spreadsheet cannot produce, because it never held the columns you would need to compute it.
Monthly
the cadence at which anticipated final cost is reviewed on a controlled project
Cost-management convention
A convention, not a measured benchmark. The distinction between a managed project and a surprised one is rarely the sophistication of the model — it is whether anybody re-runs it before the number has already happened.
Cost control is the delivery-phase ledger, kept from the owner’s chair: what each budget line was originally given, what has moved in or out of it, what has been awarded against it, what the variations will do to it, and what is genuinely left to spend.
Almost every project runs a version of this in a spreadsheet, and almost every version answers the wrong question. It reports what has been INVOICED. The question a cost manager is actually being asked — in the meeting, on the spot — is what is left, and that is a different number arrived at a different way.
A budget line is not a number. It is a small ledger, and every column on it is sourced from a different place, arrives at a different time, and is trusted to a different degree. Collapse them into one figure and you have not simplified the problem — you have deleted the information that made it solvable.
This is the model Zepth keeps. It is worth reading as a checklist against whatever you run today, because the columns you are missing are precisely where your surprises come from.
| Column | Where it comes from | What it tells you |
|---|---|---|
| Original budget | The approved budget at sanction. Frozen. | What you were given. It should never move — if it does, you have lost your baseline and with it any ability to explain the variance. |
| Transfers in / out | Approved virements between lines. | Movement between packages, each with an approval behind it. The audit trail for why a line no longer matches its sanction. |
| Current budget | Original ± transfers. | What this line is authorised to spend today. Calculated — never typed. |
| Committed | Awarded contracts and purchase orders. | Money you are contractually obliged to pay, whether or not an invoice has arrived. The gap between this and “actual” is where most cost reporting quietly lives. |
| Variations | Change orders, by status: anticipated, pending, approved, disputed. | The forecast’s live edge. Approved variations move the commitment; the others move the exposure, which is why status discipline is not clerical. |
| Direct costs | Spend with no PO behind it — day works, statutory fees, utilities, owner-supplied materials. | Real money that the classic budget-versus-contracts spreadsheet omits entirely, because it only ever looked at contracts. |
| Actual | Certified and invoiced to date. | What has been paid or is payable. The number every legacy system reports — and the one that tells you least, because it is the furthest behind. |
| Unawarded balance | Current budget − committed − approved variations − direct costs. | The scope still to be bought — and therefore the only money you can still make decisions about. Calculated. |
| Anticipated final cost | Committed + variations (by status) + direct costs + unawarded balance. | Where this line lands if nothing changes. The number the board is actually asking for when it asks how the project is going. Calculated. |
A budget line answers “what may I spend?”. It does not answer “whose money is it?” — and on a funded project those are different questions with different consequences.
Equity, senior debt, a government grant, a tenant contribution, a specific loan tranche: each has its own conditions, its own drawdown schedule, and its own idea of what it may be spent on. A grant with an eligible-cost definition will not pay for the item you charged to it, and you will find that out at the audit, long after the money left.
So the ledger carries the funding source alongside the budget, and the drawdown alongside the spend. It is the difference between knowing you are 60% through your budget and knowing you are 60% through your budget but 85% through the tranche that was funding it — which is not a reporting nuance. It is a cashflow event with a date on it.
The single most informative event in a project’s cost life is the award, and most systems record it as a procurement fact rather than a cost one.
When a package is awarded, the estimate meets the market. The buy-out variance — what you budgeted for that scope against what you actually contracted for it — is knowable that day, for that package, with no forecasting and no judgement required. It is the cleanest signal in cost management, and it is available months before anything else tells you the same thing.
A project that captures buy-out variance at every award knows by the third or fourth package whether its estimate was systematically optimistic, and has the rest of the programme in which to do something about it. A project that waits for invoices learns the same thing at handover, when there is nothing left to buy and no decision left to make.
“Every award is a measurement. A project that does not read them is choosing to learn the same lesson later, at full price.”
Ask a project how it is doing on cost and you will be told a percentage of budget spent. It is close to meaningless. Spend is a lagging measure of decisions taken weeks or months earlier; by the time it moves, the decision it reflects cannot be revisited.
The unawarded balance is the opposite. It is what remains of the current budget once you subtract everything already committed, every variation approved, and every direct cost incurred — and it is therefore the only part of the budget you still have any influence over. It is scope not yet bought, at prices not yet fixed, with alternates still available.
That makes it the right number to manage, and the right number to defend. A line with a healthy percentage of budget remaining but a negative unawarded balance is not healthy — it is a line that has already spent money it has not yet been invoiced for, and the invoices are coming.
A variation is not a single event. It moves through states, and each state has a different claim on the forecast.
Anticipated. Somebody has issued an instruction, or the design has moved, and a cost will follow — but nothing has been priced. It belongs in the forecast as exposure, not as commitment. Projects that keep this column empty are not projects with no anticipated variations; they are projects that are not looking.
Pending. Priced and submitted, awaiting a decision. This is the column that ages, and ageing here is the leading indicator of a dispute: a variation that has been pending for four months is not a cost problem yet, but it is on its way to becoming a claim. The remedy is the approval workflow, not the cost report.
Approved. It moves the commitment and the current budget’s relationship to it. From this point it is not a forecast — it is money, and it belongs to the ledger rather than the exposure.
Disputed. Held, quantified, and visible — not netted off and not quietly forgotten. A disputed variation excluded from the forecast is a lie of omission that will be corrected by someone else, at a time of their choosing.
The classic cost spreadsheet compares the budget to the contracts. It is a clean, comprehensible model, and it has one flaw: a material share of what a project actually spends never passes through a contract at all.
Day works. Statutory and authority fees. Utilities and temporary services. Owner-supplied materials. Insurance and testing. Consultant time billed against the project. None of it has a purchase order, and so none of it appears — which means the report reconciles beautifully and is wrong by the whole amount.
Practitioners will tell you this runs somewhere in the range of 10–15% of project spend. We are labelling that as what it is: an estimate from people who do this work, not a measured benchmark, and we can find no defensible published figure for it. Which is itself the point. A category of spend large enough to decide whether a project lands on budget is one that nobody has bothered to measure — because in most systems, it was never a column.
None of the mathematics on this page is difficult. Any cost manager could compute an anticipated final cost by hand. The reason projects still overrun is not that the sum is hard — it is that the inputs arrive too late to act on.
Cost is COMMITTED at decision moments: an instruction on site, an award, a design change accepted in a meeting. It is REPORTED at invoice moments, which are weeks or months downstream. Every legacy cost system reports the second and calls it control. What it is measuring is history.
The whole purpose of holding the ledger in the same system as the instructions, the purchase orders and the invoices is to shorten that gap — so that a site instruction with a cost consequence lands in the forecast on the day it is issued, not in the month-end after the claim arrives. The industry-wide figures — projects running up to 80% over budget, roughly one in ten of a contract’s value returning as change orders — are not made of catastrophes. They are made of decisions that were perfectly manageable on the day nobody costed them.
Cost Control is the delivery-phase ledger, from the owner’s chair. It is deliberately distinct from the two disciplines it is most often confused with, and the distinction is worth being blunt about.
Zepth Vector is the buying machinery — how a commitment gets raised, tendered, awarded, matched and paid. It feeds this ledger; it is not this ledger. Zepth Edge is the asset’s investment case — what capital the asset needs, over years, and whether it is funded. It precedes this ledger and it outlives it.
One record, three lenses. The purchase order that Vector raises is the commitment this page tracks and the capital that Edge sanctioned. It is entered once.
Every budget line carries the full ledger — original, transfers, current, committed, variations by status, direct costs, actual, unawarded balance, anticipated final cost — with source of funds attached, and with the three calculated columns calculated rather than entered.
Awards write back to the budget automatically, so buy-out variance appears at the award rather than at the audit. Site instructions with a cost consequence land in the forecast as anticipated variations on the day they are issued. Direct costs have a column, because they have a cost. And the anticipated final cost is a live number rather than a monthly artefact — which means the monthly review is a review, not a reconstruction.
The question “how much is left?” has an answer that is not the amount not yet invoiced.
Buy-out variance is known at the award, with the rest of the programme still available to respond — rather than at handover, with nothing left to buy.
Direct costs are in the ledger, so the report is not silently wrong by the one category nobody has ever measured.
A cost consequence enters the forecast on the day it is decided, rather than in the month-end after the claim.
Source of funds sits beside the budget, so a fully-drawn tranche behind a half-spent budget is a fact rather than a surprise.
Original, transfers, current, committed, variations by status, direct costs, actual, unawarded balance and anticipated final cost — per line, with the calculated columns calculated.
Equity, debt, grant or tranche carried alongside the budget, with drawdown tracked against spend and eligible-cost conditions attached where they exist.
Contracts and POs write back to the line, so committed cost is real from the moment of award and buy-out variance is captured with it.
Anticipated, pending, approved and disputed — each weighted correctly in the forecast, with ageing on the pending column.
Day works, statutory fees, utilities and owner-supplied materials booked to the budget line without a purchase order, because they belong to the cost even though they never met a contract.
Recomputed as the inputs move, so the monthly cost review is a review rather than a month of reconstruction.
Original budget frozen at sanction, and never editable afterwards. Whose money it is travels with it, because a grant, a tranche and equity do not spend alike.
Awards and purchase orders write back to the budget line. The commitment exists from the award, not from the invoice — which is the entire difference between control and bookkeeping.
Budgeted scope against contracted scope, package by package. The cleanest signal in cost management, and the earliest.
Anticipated, pending, approved, disputed — each with its own claim on the forecast. A pending variation that is ageing is a dispute forming, and it is visible long before it is a claim.
Day works, fees, utilities, owner-supplied materials. No purchase order, but unmistakably money — and a column, so that the report is not wrong by the whole amount.
The AFC says where the line lands. The unawarded balance says how much of that you can still change. One is the diagnosis; the other is the treatment.
A site instruction, an RFI response or a design change with cost implications is surfaced as an anticipated variation on the day it is issued — while it is still a decision rather than a claim. The agent proposes; the cost manager decides what enters the forecast.
“Your last four awards came in 6–11% above budget, all mechanical.” A pattern in the estimate, visible after four packages rather than forty.
Ranked by the trajectory of the anticipated final cost rather than by the size of the line — because the biggest line is rarely the one about to hurt you, and the one about to hurt you is usually visible weeks earlier in its unawarded balance.
Pending variations that are ageing, and disputed ones that have gone quiet. Both are cost, and both look like nothing until they arrive together.
The engineer’s judgment stays in charge; the AI removes the latency and the blind spots.
Budget versus committed versus anticipated final cost, per line and rolled up to the project. The unawarded balance, which is the only part of the budget still open to a decision. Buy-out variance by package, from the award rather than the invoice. Variation exposure by status, with the pending column aged. Direct costs, which most systems cannot report because they never held them. Drawdown against each source of funds, so a fully-drawn tranche behind a half-spent budget is visible while it is still a cashflow question rather than a cashflow event. And the movement in anticipated final cost month on month, which is the single report that answers what the board is actually asking.
The current budget less everything already committed, less approved variations, less direct costs incurred. It is the scope still to be bought — and therefore the only part of the budget you can still make decisions about. It is the number to manage, because percentage-of-budget-spent is a lagging measure of decisions you can no longer revisit.
Read the full answerThe budget is what you were authorised to spend. The anticipated final cost is where the line actually lands if nothing changes: committed cost, plus variations weighted by status, plus direct costs, plus the unawarded balance still to be bought. The budget is a permission; the AFC is a forecast. A project that reports only the first is reporting only what it was allowed to do.
Read the full answerBecause a budget answers what may be spent, and a funding source answers whose money it is — and each has its own drawdown schedule and its own conditions on what it may be spent against. A grant with an eligible-cost definition will not pay for an item charged to it in error, and that is discovered at the audit. Being 60% through the budget but 85% through the tranche funding it is a cashflow event with a date on it, not a reporting nuance.
Read the full answerSpend with no purchase order behind it: day works, statutory fees, utilities, owner-supplied materials, testing, consultant time. They are missing because the classic cost model compares the budget to the contracts, and direct costs never met a contract. Practitioners put this at somewhere around 10–15% of project spend; we can find no defensible published figure, which tells you how thoroughly it has been overlooked.
No. Budgeting sets the number and manages it against a forecast — that is Budget Management, and it is where transfers between lines are governed. Procurement raises the commitment and pays it — that is Zepth Vector. Cost control is the delivery-phase ledger in between: what has been committed against each line, what the variations will do to it, and what is genuinely left. Same record, different lenses.
Monthly is the working convention, and it is a convention rather than a benchmark. But the cadence matters less than the latency: an AFC recomputed monthly from inputs that arrive monthly is still reporting last month. The point of holding instructions, awards and invoices in the same system as the ledger is that the forecast moves on the day the decision is taken, so the monthly review is a review rather than a reconstruction.
Related modules
Related answers
Zepth is the construction project delivery platform — it runs construction, procurement and asset management on one record, and does the work: reading the drawings, reviewing the submittals, matching the invoices and flagging the risks, with a human sign-off on anything consequential.
A short, tailored walkthrough on your real workflow — no generic demo.