Job Costing Methods: How Losing Projects Hide Inside Healthy-Looking Margins

Your firm-wide margin reads 34%. It has read roughly 34% for a year. So you assume every project is broadly healthy. That assumption is the trap. The right job costing methods exist to break it — to show you which projects earned that 34% and which ones quietly burned cash while the winners covered for them.

Here is the uncomfortable part. Aggregate margins are an average, and averages hide their extremes. A project losing 20% and a project earning 50% blend into a number that looks reasonable. You never see either. You see the middle, and you make decisions on the middle.

Why the aggregate number lies to you

Averaging across a portfolio is exactly how loss-makers stay invisible. When one profitable job subsidizes a bleeding one, the total looks stable — and stability reads as health.

Say you run ten projects. Eight earn a 40% margin. Two lose 15% each. Blended, you land near 34%. The dashboard is calm. Meanwhile two clients are consuming your best people at a loss, and nothing on the P&L tells you which two.

This is the core failure of costing at the portfolio level instead of the project level. Project accounting exists precisely to track revenue and cost per engagement, not per month. Without it, you cannot answer the only question that matters: which work makes money, and which work costs you to keep?

The dangerous version isn't a job that loses money loudly. It's the job that looks fine because a better one is paying its rent.

Job costing mistakes accumulate quietly for months

Bad job costing rarely fails on day one. It drifts. Small allocation errors compound into job-level data that looks reasonable but misrepresents each project.

The usual culprits are boring and cumulative:

1. Unlogged hours. A senior consultant spends three hours unblocking a stuck project and never codes them. That project's real cost is understated every week.
2. Flat labor rates. You cost every hour at one blended rate instead of the true loaded cost per person. See how to build a real labor cost rate — the fully burdened number, not the salary divided by 2,080.
3. Unallocated overhead. Software, admin time, and non-billable coordination get dumped into a general bucket and never hit the jobs that caused them.

None of these throw an error. They produce a number. The number just isn't true. As NetSuite's project accounting guidance puts it, accuracy depends on capturing costs at the transaction level and tying them to the specific project — otherwise you're estimating, not costing.

Months of small estimation errors don't cancel out. They settle onto the projects that get the least scrutiny — the quiet, mid-sized ones nobody flags.

The three job costing methods, and when each fits

There is no single correct method. There's the method that matches how your firm actually delivers work. Pick wrong and the data will mislead you cleanly.

• Actual costing. You assign real hours and real costs to each job as they happen. Most accurate. Requires disciplined time capture. Best for firms where labor is the dominant cost and hours vary widely by project.
• Standard (predetermined) costing. You cost jobs at expected rates set in advance, then track variance against actuals. Faster, cleaner for quoting, and useful when your work is repeatable. The risk: if standards drift from reality, so does every job's margin.
• Normal costing. A hybrid — actual direct costs plus a predetermined overhead rate applied by a driver like billable hours. Practical for most professional services firms, where direct labor is measurable but overhead is messy to trace directly.

Scoro's project accounting primer walks the mechanics well. The strategic point is simpler: the method determines what you can see. Standard costing surfaces variance. Actual costing surfaces truth but demands rigor. Choose based on the decision you need the data to support.

The metrics that expose the subsidy

A costing method only helps if it feeds the right metrics. Track two things per project and the hidden loss-makers stop hiding.

First, project-level margin — revenue minus fully loaded cost, per engagement, not blended. Our guide to project margin calculation breaks down the mechanics, and the margin calculator gives you a fast read. Watch contribution margin alongside gross margin: contribution tells you whether a job covers its own variable cost before overhead ever enters the picture.

Second, utilization by project and by person. HBR's research on professional service firms is blunt that leverage and utilization drive profitability more than headline rates. A job can bill well and still lose money if your most expensive people are staffed to it below their cost-recovery point. Our utilization rate hub and consultant utilization guide cover this in depth.

NetSuite's KPI breakdown for consulting firms lists the full set worth watching. Start with margin per project and utilization. Those two catch the subsidy.

What to fix this quarter

You don't need a system overhaul to stop making decisions on averages. You need three moves.

1. Rank every active project by margin, worst to best. The bottom of that list is where your subsidy lives. Most owners have never seen this ranking.
2. Fix your cost inputs before your method. A perfect costing method on wrong labor rates still lies. Get the loaded rate right, then capture the hours.
3. Review at the project level monthly, not the firm level quarterly. The AICPA's practice management resources reinforce that firm health is the sum of engagement health — you have to look one layer down.

For the deeper mechanics, our project costing hub and the broader professional services finance library cover related pieces like work-in-progress accounting and unbilled revenue that shape the true picture.

The takeaway

A healthy aggregate margin is not evidence that your projects are healthy. It's evidence that your winners are strong enough to hide your losers. Job costing methods exist to end that guessing — to show you, per engagement, who's paying and who's being paid for.

CentSight is the intelligence layer on top of QuickBooks and your bank, built to surface those project-level signals before they compound. Connect your accounting software in a few minutes. If it confirms what you already know, you're out $95. If it doesn't, it just found the loss-making job your best client was quietly funding.