Your fixed-fee project can have money left in its budget and still be heading for an overrun. The missing number is the cost of the work your team hasn't done yet. Put a fresh estimate of that work beside actual costs and approved fees each week, so the project manager can decide what to change while the job is still open.
We helped one retail engineering firm report about $1.98M in annual margin improvement through work on unpriced changes, repeat drawing issues and earlier budget decisions. That total combined $1.2M a year in priced and billed changes with a $784K estimated annual benefit from less rework. As part of the engagement, we helped managers review pilot-store budgets weekly. One manager who'd discovered an overrun too late in a monthly report acted before closeout on a later job.
Here's how to make that earlier decision possible in your own review. In the worked example below, a $100,000 fee and $60,000 spent look comfortable until you include the work still needed to finish. You'll build one forecast that shows the remaining contribution, keeps hoped-for fees separate and gives a named manager a decision to make.
Start with the fee you can use and the costs you mean
Let's say you're reviewing a fixed-fee retail-design project. The agreed fee is $100,000, with no approved amendments. Finance has recorded $60,000 in delivery costs through Friday. The team estimates another $25,000 to complete the agreed work.
The forecast total cost is $60,000 plus $25,000, or $85,000. Subtract that from the $100,000 fee and the forecast delivery contribution is $15,000, or 15% of the fee. That's the amount left after the delivery costs you've included, before any other firm costs. Use that name and cost definition consistently so a project contribution doesn't become a claim about net profit.
Finance should set the basis before the first review. For this example, delivery cost means labor at agreed internal cost rates, including the labor burden finance specifies, plus direct project expenses. It excludes other firm overhead. Use cost rates for effort, not client billing rates, because you're estimating what the remaining work will cost to deliver.
A common cutoff matters just as much as the formula. If Friday's hours haven't posted, add an identified estimate for that incurred work. Remove that estimate when the actual entry arrives, so the same hours aren't counted twice. Separate work already done from work still to do. A low actual-cost total is misleading when it simply leaves out this week's effort.
Ask the team what remains, not how much budget is left
The remaining $25,000 needs a work estimate behind it. Ask each discipline lead and the review owner what they still need to complete the current scope, then price those hours on the same basis as the actuals. Include coordination, checking, corrections, expected resubmittal work and closeout where the job requires them.
| Remaining work in the example | Estimate | Basis to confirm |
|---|---|---|
| Finish design and coordination | $12,000 | Remaining tasks and hours agreed with discipline leads |
| Technical review and planned corrections | $8,000 | Review effort and a correction allowance confirmed by the review owner |
| Issue and closeout | $5,000 | Remaining issue tasks and direct expenses confirmed by the project manager |
| Total cost to complete | $25,000 | The same cutoff and cost basis as actual costs |
The table is a teaching example. In your project, keep the hours, rates, source record and confirming owner behind each amount. A blank estimate means the work hasn't been estimated. It doesn't mean the work will cost nothing.
Subtracting spent cost from the old budget won't tell you what's left to do because the original plan may have missed a review round or a client change. Ask for the remaining tasks first, then compare their cost with the plan. That comparison tells the manager where the forecast changed and who can explain it.
This review gives you a project decision before the remaining fee is consumed. If your larger goal is to grow profit across the firm, the next question is which recurring loss deserves the team's attention first.
If you’re an engineering firm leader and want to grow your firm without finding project overruns after the work is finished, we’ll identify the biggest bottleneck standing in your way, build the plan to remove it, and implement it in eight weeks* or less - guaranteed or we continue to work for free and you don’t pay.
Find my bottleneck* Eight-week delivery, or twelve weeks when substantial data foundations are required, as agreed before work begins. If a covered stage fails its agreed benchmarks, we provide up to four additional weeks of remedy work at no charge, then waive or refund that stage’s fee if it still fails. Successfully delivered stages remain payable. Agreed client-caused delays pause the clock. Delivery guarantee terms.
Change the forecast when the work changes
Now the review owner identifies another $10,000 of coordination and correction work. It wasn't included in the original $25,000 estimate. Actual cost is still $60,000 because the team hasn't done that new work yet, but the forecast has already changed.
| Forecast measure | Before the new work | After the new work |
|---|---|---|
| Approved fee | $100,000 | $100,000 |
| Actual delivery cost | $60,000 | $60,000 |
| Remaining delivery cost | $25,000 | $35,000 |
| Forecast total cost | $85,000 | $95,000 |
| Forecast delivery contribution | $15,000 | $5,000 |
| Contribution as a share of approved fee | 15% | 5% |
The contribution has fallen by $10,000 before the accounting report shows another dollar spent. That's why the review owner needs to update the estimate when the work becomes known. Waiting for the timesheet would leave the project manager less time to respond.
The manager also has a $12,000 amendment request awaiting client approval. Keep it outside the base fee. If the client approves it for work already included in the revised $35,000 remaining cost, the fee would become $112,000 and the forecast contribution would become $17,000. If the amendment adds more work, include that extra delivery cost in the scenario too. Otherwise the forecast credits the fee but forgets what you must deliver for it.
Label the pending amount with its source, owner and next decision date. Keep any authorized work's cost in the forecast even while its fee remains pending, so a delay in approval doesn't make the job look healthier than it is. Use the scope-change triage method when you need to trace the request and prepare a commercial decision.
The project manager can now investigate the new $10,000: required corrections, an added client request, or work already counted elsewhere. Each cause changes the action. A duplicate needs a corrected forecast. A client request needs a scope and fee decision. Required corrections need a feasible delivery plan and a leader who accepts the cost. Removing necessary review from the estimate wouldn't remove the work.
Record the chosen action, its owner and its due date. For this example, the manager asks the principal to decide on the amendment and the review owner to confirm the revised work plan before committing optional additional scope. If required work must continue, the principal records that choice with the $5,000 base contribution visible.
Use AI to gather evidence, then let owners approve the forecast
Start with the project system your team already uses. Deltek's Vantagepoint revenue-forecast documentation describes entering estimate-to-complete revenue across a project's structure. Check your existing planning setup before building another report, and keep that revenue view separate from the remaining delivery cost you're calculating here.
If the estimate is already current but managers won't act on it, another collection tool won't solve that decision gap. If the recurring delay is finding changes across emails, review logs and project records, an AI-assisted collection step is worth testing. Its job is to bring possible missing work to the person who can confirm it.
Give that step the permitted project records, current scope and amendments, latest task estimates, review log and correspondence since the last cutoff. Keep project IDs, document versions and access permissions attached. The system should return candidate work items with a source passage, the affected task, any stated hours, the proposed owner and the reason the item may change the estimate. Missing or conflicting information stays unresolved until an owner checks it.
For the example's new review work, a candidate output could read:
Project EX-01. Source: review log R7, item 4. Coordination and correction tasks are absent from the current remaining-work list. Hours: not supplied. Owner: review lead. Action: confirm tasks and effort, then check whether they overlap the existing correction allowance. Status: awaiting owner review.
The review lead supplies the effort and confirms the overlap. Finance's cost rates turn the approved effort into dollars, and a spreadsheet or fixed calculation updates the total. Keep the AI output in a staging list until that approval happens. The model shouldn't invent hours, approve an amendment or overwrite the project's financial record.
Test this on completed jobs using only records available at each historical cutoff. Ask the project manager and review owner to mark which items should have changed the forecast, then compare the proposed output with that reference. Count missed material work, duplicates, unsupported estimates, wrong-project matches and the time needed to check each output. A plausible summary is no help if its omissions leave the forecast wrong or its review takes longer than collecting the inputs directly.
Then run one live project beside the existing process. Agree the acceptable error and review-effort limits before the trial, keep the existing forecast authoritative and expand only if the output meets those limits. An uncertain source goes to its owner. It doesn't become a guessed cost. Give the internal system owner the rules and test examples so later changes can be checked before use.
Leave the weekly review with one decision someone owns
Copy this card for one active fixed-fee project. The project manager assembles it, discipline and review leads confirm remaining work, and finance confirms the fee and cost basis. The worked column shows the revised example with the amendment still pending.
| Forecast card field | Blank entry | Worked example |
|---|---|---|
| Project, review and cutoff | [Project ID, review date, cost cutoff] | EX-01, Monday review, Friday cutoff |
| Cost basis and owner | [Included labor burden, direct expenses, exclusions, finance owner] | Finance-approved labor costs and direct expenses, other firm overhead excluded |
| Agreed fee and approved amendments | [Base fee, approved additions, approval links] | $100,000 + $0 approved amendments |
| Pending fee changes | [Amount, source, approval owner, decision date] | $12,000, change request C3, principal to follow up today |
| Actual cost and unposted work | [Posted cost, separately identified incurred estimate] | $60,000 posted, $0 unposted after confirmation |
| Remaining work | [Tasks, hours, cost rates, expenses, confirming owners] | $12,000 design + $8,000 planned review + $5,000 closeout + $10,000 new work |
| Forecast total delivery cost | [Actual + unposted incurred estimate + remaining cost] | $95,000 |
| Base contribution and percentage | [Approved fee minus forecast cost, divided by approved fee] | $5,000, or 5% |
| Change from prior review | [Amount, cause, source and unresolved assumptions] | $10,000 more cost from R7, review lead confirms no overlap |
| Separate approval scenario | [Pending fee and any associated extra costs] | $17,000 contribution if C3 is approved and its work is already in the cost estimate |
| Decision, owner and deadline | [Action, accountable owner, due date] | Principal decides on C3 today, review lead confirms work plan before optional scope is committed |
| Next review and eventual check | [Next cutoff, forecast version, closeout comparison] | Next Friday cutoff, save v2, compare forecast with final cost |
A weekly meeting is the regular checkpoint. A material new request or review finding should reach the owner when it arrives, because waiting for the next meeting can consume the time available to act. If the record is complete but an alert doesn't reach its decision-maker, use the project-manager action guide to inspect that handoff.
Keep each forecast version. At closeout, compare its total-cost estimate with final delivery cost and explain the difference using scope changes, missing work or wrong effort assumptions. That gives the team something specific to improve in the next estimate. Also record how long material changes waited for a decision, because a more accurate forecast only helps the live job if someone uses it in time.
For your next review, complete this card for one project and leave with one named decision before further optional work is committed. You'll have a cost-to-finish estimate the team can explain, a base contribution that doesn't depend on an unsigned amendment and an owner for the action that follows.
That makes one project easier to manage. If your goal is stronger profit across the firm, use the repeated causes across projects to decide where to investigate first. The biggest limit on growth may sit in pricing, delivery, capacity or another part of the business, so the broader engagement starts with your goal and the evidence behind it.
If you’re an engineering firm leader and want to grow your firm without finding project overruns after the work is finished, we’ll identify the biggest bottleneck standing in your way, build the plan to remove it, and implement it in eight weeks* or less - guaranteed or we continue to work for free and you don’t pay.
Find my bottleneck* Eight-week delivery, or twelve weeks when substantial data foundations are required, as agreed before work begins. If a covered stage fails its agreed benchmarks, we provide up to four additional weeks of remedy work at no charge, then waive or refund that stage’s fee if it still fails. Successfully delivered stages remain payable. Agreed client-caused delays pause the clock. Delivery guarantee terms.