The most important number in a construction cost report isn't what's been spent. It's the forecast of what the project will cost when it's finished. That's the number a board approves funding against, the number a lender watches and the number that decides whether contingency is enough. Yet on many projects the forecast is simply the budget plus approved changes, updated only when something goes obviously wrong. By then, the options for fixing it are limited.
This guide explains estimate at completion (EAC) and estimate to complete (ETC), the main ways to calculate them, why a single forecast is never enough, and how owners can build one they can defend.
The terms
- Budget at completion (BAC): the approved budget for the work.
- Actual cost (AC): what's been spent, or more precisely, what's been incurred, to date.
- Estimate to complete (ETC): what the remaining work will cost from today.
- Estimate at completion (EAC): actual cost to date plus the estimate to complete. It's the forecast final cost.
- Variance at completion (VAC): budget minus EAC. A negative number is a forecast overrun.
The whole challenge, and most of the judgment, is in the ETC. Actual cost is a fact. The cost of the remaining work is a judgment, and the method used to make that judgment changes the answer.
Two ways to forecast: bottom-up and index-based
There are two broad approaches, and good cost control uses both.
Bottom-up forecasting builds the ETC from the remaining work, package by package. For each trade or cost code, it starts from the committed contract value, adds approved and pending changes, estimates anticipated changes that haven't been priced yet, and allows for known risks. This is the forecast most owners and cost consultants produce each month.
Index-based forecasting uses earned value performance to date to project the rest. If the project has been getting 91 cents of work for every dollar spent, the index-based forecast assumes something similar will continue. Our guide to earned value management for owners explains the underlying measures.
The two methods check each other, and neither is enough on its own. A bottom-up forecast that's much lower than every index-based forecast is assuming performance will suddenly improve. That might be justified, but someone should be able to say why.
Building a bottom-up forecast
For each cost line, a credible forecast should account for:
- Committed cost: the value of awarded contracts and purchase orders.
- Approved changes: executed change orders.
- Pending changes: change requests received but not yet agreed, at a realistic value, not the contractor's opening number or zero.
- Anticipated changes: changes the team knows are coming but that haven't been submitted yet, such as a design revision in progress or a known site condition.
- Uncommitted work: packages not yet awarded, at current market pricing rather than the original estimate.
- Risk allowance: money held for specific, identified risks that haven't happened yet.
View the data
| Component | Amount |
|---|---|
| Original budget | $20.0M |
| Approved changes | +$0.8M |
| Pending changes | +$0.5M |
| Anticipated changes | +$0.4M |
| Risk allowance | +$0.6M |
| Forecast final cost | $22.3M |
The step most often skipped is anticipated changes. On programs we've supported, the forecast that surprised a board was rarely caused by a change nobody saw coming. It was usually caused by changes the site team knew about for months that never made it into the report because nobody had priced them yet.
Index-based EACs: the main formulas
The U.S. Government Accountability Office's Cost Estimating and Assessment Guide describes the general form: actual cost to date, plus the remaining budgeted work divided by a performance index. The index is what changes between methods.
- CPI-based EAC. Uses the cost performance index. GAO notes this assumes the contractor will keep operating at the same efficiency, and that it typically produces the lowest EAC, so it's best treated as a best case.
- SPI-based EAC. Uses the schedule performance index, assuming schedule slippage drives cost. GAO notes it's more accurate early in a project than later.
- Composite (CPI × SPI) EAC. Assumes that recovering lost schedule will cost extra, so the two effects compound. GAO describes it as usually producing the highest EAC, a worst-case predictor.
- Weighted EAC. Blends cost and schedule performance into one index. GAO describes a 50/50 weighting of CPI and SPI as statistically the most accurate.
A worked example
Take an illustrative $20 million project at the halfway point of its schedule. The earned value data at the reporting date:
- Budget at completion: $20.0 million
- Planned value: $9.0 million
- Earned value: $8.0 million
- Actual cost: $8.8 million
That gives a CPI of 0.91 and an SPI of 0.89. The remaining budgeted work is $12.0 million. Running it through each method:
- CPI-based: $8.8M + $12.0M ÷ 0.91 = $22.0 million
- SPI-based: $8.8M + $12.0M ÷ 0.89 = $22.3 million
- Weighted 50/50: $8.8M + $12.0M ÷ 0.90 = $22.1 million
- Composite CPI × SPI: $8.8M + $12.0M ÷ 0.81 = $23.7 million
Now suppose the contractor's monthly report forecasts a final cost of $21.0 million. Every index-based method says that's optimistic. The spread between the methods, roughly $1.7 million on a $20 million project, is itself useful information. It shows how much the answer depends on whether the schedule slippage gets recovered and what that recovery will cost.
View the data
| Method | EAC |
|---|---|
| Contractor forecast | $21.0M |
| CPI-based (best case) | $22.0M |
| Weighted 50/50 | $22.1M |
| SPI-based | $22.3M |
| CPI × SPI (worst case) | $23.7M |
The reality check: to-complete performance index
The to-complete performance index (TCPI) asks a simple question: how efficiently does the remaining work have to be done to hit a given target? To finish on the original $20 million budget, TCPI is the remaining work ($12.0 million) divided by the remaining money ($11.2 million), or about 1.07.
In plain terms, the project would have to deliver $1.07 of work for every dollar spent from now on, after delivering only $0.91 so far. GAO's guidance is direct: when the TCPI is higher than the CPI, the forecast assumes productivity will improve, and analysts should look for evidence that it will. Without that evidence, the budget isn't a forecast. It's a hope.
Why a range beats a single number
GAO calls it a best practice to develop more than one EAC, so management can see the likely range of final cost. It also recommends preparing EACs below the project level, so that poorly performing packages aren't hidden by good ones elsewhere.
For owners, presenting a range has two practical benefits. It's more honest, because the final cost really is uncertain. And it changes the board conversation from "is the number right?" to "what's driving the range, and what are we doing about it?" That's a much more useful discussion.
Don't forget time-related costs
A forecast built line by line from trade packages can miss the costs that grow with time rather than scope. If the schedule is slipping, the cost forecast should reflect it. The usual items:
- Contractor general conditions: site supervision, temporary facilities, equipment rental and insurance that run for as long as the job does.
- Owner-side costs: project management, consultants' contract administration, security, interim leases and financing carry.
- Escalation: work bought later is generally bought at a higher price.
The schedule and the cost forecast should be read together every month. A forecast that holds steady while the completion date moves out three months is almost certainly missing something.
Who should own the forecast
The contractor's forecast covers the contractor's scope, and it's useful. But it isn't the owner's forecast. The owner's total project cost also includes design fees, permits, owner-supplied equipment, furniture, IT, moves and its own staff. The owner, or its representative, should own a forecast that covers everything, and use the contractor's numbers as one input to it.
Common forecasting mistakes
- Budget plus approved changes. The forecast ignores pending and anticipated changes until they're signed.
- Billing treated as cost. Actual cost is taken from contractor invoices, which may run ahead of or behind the work actually done.
- Contingency treated as spent or unspent. Contingency is shown as fully available until the moment it's needed, instead of being drawn down against identified risks.
- No check against performance. The bottom-up forecast is never compared with an index-based EAC.
- Forecasts that only move at milestones. The number stays flat for months, then jumps.
What owners should do
- Require a monthly bottom-up forecast by cost line, showing committed, approved, pending, anticipated, uncommitted and risk amounts.
- Run at least two index-based EACs alongside it: a CPI-based best case and a composite worst case.
- Check the TCPI against the CPI every month. If the forecast needs a big jump in productivity, ask what will cause it.
- Report a range to the board, with the main drivers of the gap between low and high.
- Forecast at package level, so a problem in one trade isn't hidden by savings in another.
The bottom line
A cost forecast is only useful if it changes before the money is spent. Owners who build bottom-up forecasts, test them against performance and report an honest range see overruns while there's still time to act. Owners who wait for signed change orders see them after the fact, when the only remaining options are finding more money or cutting scope.
Want an independent forecast of where your project's final cost is heading? Book a 30-minute call with Benjamin R. Consulting.
Sources: U.S. GAO, Cost Estimating and Assessment Guide: Project Future Performance; Benjamin R. Consulting, Earned value management for construction owners, explained simply. The worked example is illustrative. This article is general information, not financial or legal advice.