All posts

Projected Cost to Complete: Catch Overruns Before Closeout

Professional header image for educational tutorial: Projected Cost to Complete: Catch Overruns Before Closeout
AI-generated header image for: Projected Cost to Complete: Catch Overruns Before Closeout

The first time many contractors learn a job lost money is at closeout, when nothing can be recovered. That is not job costing; it is an autopsy. Overruns rarely arrive all at once. Material price swings, labor shortages, scope changes, weather delays, and productivity shortfalls each add a little, and small variances in labor or materials compound across multi-phase or tightly bid work until the margin is gone.

A projected cost to complete changes that timeline. By rebuilding the forecast at set milestones from actuals, commitments, and remaining scope, you can see trouble while there is still room to react. Thresholds make the forecast useful: when a variance crosses the line, someone has to have a conversation instead of waiting for the surprise.

This tutorial walks through why small variances surface late, what a projected cost to complete actually measures, how to choose milestones that give you time to act, how to set thresholds that force a conversation, how to match report formats to contract type, what that trigger conversation sounds like, and how to make the monthly cadence stick.

Why Small Variances Show Up Late in Job Costing Reports

Job costing exists to give you decision time. A forecast that only lands at closeout is a report, not a warning.

A cost-to-complete session for subcontractors names the overrun sources that job cost data registers long before an invoice does: material price movement, labor availability and productivity, scope changes, weather delays, and productivity drift.

Closeout review is structurally too late. By the time final invoicing reconciles actuals, the crew has demobilized, change-order notice windows have often closed, and the only lever left is absorbing the loss. Cost control research makes the same point: project accounts are generally hard to interpret until a project is finished, and by then it is too late to influence the outcome.

That is the trap in many job costing programs. Labor and material codes are applied accurately, the cost detail per project is rich, and nobody reads the trend line until the job is closed and the margin is final.

None of this is a data problem. It is a timing problem, and it is fixable.

What a Projected Cost to Complete Actually Measures

If those variances are the signal, the forecast is what reads it. A projected cost to complete rests on three inputs.

Actuals are costs already coded to the job. Committed costs are subcontracts, purchase orders, and material agreements signed but not yet invoiced. Remaining scope is the work still to be performed, priced at projected quantities and rates. Building the forecast from all three is Crew & Cost's working method, not a formula mandated by the CFMA session referenced above.

The arithmetic is simple enough to rebuild by hand:

  • Projected cost to complete = commitments + remaining scope

  • Projected gross margin at completion = contract value − actuals − commitments − remaining scope

Committed costs are the blind spot. A report built only from approved invoices can show a healthy margin on a job that is already underwater once open purchase orders and subcontract obligations are counted. Exposing that gap is the entire point of an early-warning forecast.

The outputs carry further than margin. Accumulated cost and progress data also feed progress billing, retainage management, lender reporting, and bonding capacity, which is why a forecast that supports how progress billing moves cash accurately is worth more than reassurance.

Choosing Milestones That Give You Time to Act

With the forecast defined, the next decision is when to run it. No cited source prescribes one milestone scheme, so choose from three families and match the trade-offs to the job.

Percent-complete triggers surface drift earliest, but they depend on defensible progress measurement, the hardest input to keep honest. A defensible claim rests on something countable: installed quantities, units placed, a signed inspection. "Roughly 60 percent done" is a guess, and guesses hide the drift the forecast exists to catch.

Phase or milestone completion triggers align with how work is already scheduled and billed, so adoption is easy. On short phases they can fire late: a phase that consumes its cost in three weeks triggers after the money is spent.

Calendar cadence, meaning a monthly forecast, matches the repeatable monthly model CFMA teaches and is the easiest to sustain across many simultaneous jobs.

The selection rule is density, not doctrine. Match milestone frequency to job duration, phase length, and billing cycle, and put the tightest checkpoints on the thinnest-margin jobs. A year-long job billed monthly may need only monthly and phase triggers; a 90-day, tight-bid job warrants checks every 25 percent. Federal billers have a further reason to align the forecast cutoff with the invoice: prompt-payment deadline rules turn date discipline into money.

Once the milestones exist, the next question is what number at each checkpoint should trigger a review.

Setting Variance Thresholds That Force a Conversation

A threshold is a conversation trigger, not a kill switch. Crossing it routes the job into a review with a named owner and a deadline. It does not stop work, freeze purchasing, or assign blame.

No cited source defines specific variance thresholds, so treat these shapes as examples: a projected margin decline of a set number of points at a milestone, remaining cost growing above a set percentage of the original estimate, or percent complete advancing while cost consumes faster. Keep each band readable against how the work is structured, since job structure drives job costing choices as much as contract language does.

Set the band to the contract. Thin-margin lump-sum work warrants narrow thresholds. Cost-plus arrangements, where variance is more recoverable, can carry wider ones.

Build the escalation path before anything fires. Decide who is notified, within what window, what evidence the review requires, and what the review is authorized to decide. An escalation with no owner and no deadline is a notification, not a control.

Finally, separate leading signals from lagging ones. Committed cost growth, productivity trends, and quantity variance move first. Invoiced cost variance lags by weeks or months, surfacing after the cheapest corrective options have already expired. Point your thresholds at the signals that still leave room to respond.

With thresholds defined, the report itself has to present the variance they are watching, and that format depends on the contract.

Match the Report Format to the Contract Type

Thresholds tell you when to look; the report format determines what you actually see. CFMA's cost-to-complete session makes the point directly: format should follow contract type, because unit cost, percent-complete, and lump-sum work each expose variance differently.

Unit cost contracts. Report cost per unit installed against estimated unit cost. That separates quantity variance from rate variance, and the distinction matters because the fix differs: installing more units than planned is a takeoff or scope problem, while paying more per unit is a purchasing or productivity problem.

Percent-complete contracts. Report earned value against actual cost consumed, and flag any widening gap between progress claimed and dollars spent. That gap is the earliest readable symptom of an overrun, showing up before committed costs fully land.

Lump-sum contracts. Report by cost code against the original estimate. With no unit rate to normalize against, unexplained growth in remaining cost is the primary signal, and multi-phase contracts demand more than one generic view can provide.

The practical consequence for job costing programs is blunt: one generic format applied across every contract type buries the exact variance the reader needs to see. Format selection belongs in the design of the forecast, not in the printing step. Match the format, and the number that reaches the trigger conversation is the one worth arguing about.

What the Trigger Conversation Sounds Like

With the report format settled, the meeting it feeds has to be structured just as deliberately.

Who belongs in the room: the project manager, the estimator or chief estimator who priced the work, and an owner or finance lead. On subcontracts, add the general contractor contact whenever the variance touches shared or adjacent scope.

What to bring: the forecast build showing actuals, commitments, and remaining scope; the delta against the original estimate by cost code; and a written assessment labeling the gap recoverable, partially recoverable, or structural. Name the baseline explicitly, since the same job can look over or under budget depending on which estimate you measure against.

Sequence recovery before outcomes. Work through resequencing, correcting a productivity problem, pursuing change-order entitlement, or re-buying material first, then state the projected result at completion. Forecasting is early risk visibility for communicating with owners, general contractors, or subcontractors, and trigger-date tracking in billing keeps that communication tied to contractual deadlines rather than memory.

Raise a projected overrun by leading with the forecast and the recovery plan, keeping the number attached to its cost codes, and presenting it as a range with a stated basis rather than a certainty. Never wait for the next scheduled meeting: the decision value of the forecast decays with every week the job runs without a response.

Making the Monthly Cadence Stick

The conversation is only as useful as the rhythm that produces it. A monthly model can run without an ERP system or anything more than a spreadsheet. Read your job costing program as automation of a routine you already run, not a prerequisite for starting.

Assign a named owner for the monthly build. That person is accountable for the deadline, not the entire estimate. Then lock a submission date against your billing cycle so the forecast and the invoice share a cutoff; a shared cutoff keeps both on the calendar instead of turning one into an after-hours side project. If month-end feels too slow for fast-moving jobs, tighten the input with weekly cost capture between builds.

Keep the record. Each month's forecast compared against the following month's actuals is your calibration file, and it is the most direct test of whether thresholds are set too tight, too loose, or about right.

Crew & Cost keeps job tracking, costs, invoicing, and closeout in one place, so the monthly forecast is assembled from live job figures rather than a month-end scramble.

Finally, resist the audit mindset. A cost-to-complete forecast is an operating routine that has to survive staff turnover and busy quarters, not a year-end reconciliation performed once under pressure. Build it to run without you, and it will still be running in your worst month.

Conclusion: Run Job Costing as an Early-Warning System

With the monthly routine established, the build sequence is straightforward: choose milestones that match job duration and billing cycle, define the projected cost to complete as actuals plus commitments plus remaining scope, set variance thresholds as conversation triggers rather than verdicts, and match the report format to the contract type.

Do not roll it out across the portfolio. Run the first forecast this month on one high-risk open job, then compare it against what actually happened. Comparing forecast against actual earns the credibility to widen the process.

Keep the internal language honest. A projected overrun is information that arrives while there is still time to act. The entire value of the forecast is the decision window it opens, not the precision of the number.

Commit to the cadence before you shop for tools.

Set the next action now: put the first trigger review on the calendar before the first forecast is built. The meeting exists before the number does, so a crossed threshold starts a scheduled conversation instead of an improvised one.