Reading Job Cost Variances Before They Become Losses

Most Variance Reports Get Filed, Not Read

Every job cost system produces one. Budget in one column, actual in the next, variance in the third, usually with red numbers where things went over. It gets printed, glanced at, and filed.

The reason it gets ignored is that on its own it does not tell you anything actionable. A code showing $12,000 over could be a disaster or completely fine, and the report gives you no way to tell which.

The First Question: Over Against What?

A variance is only meaningful against the same percentage of work. If a cost code is 70% spent and 70% complete, there is no variance — there is a job progressing normally.

So before reading any variance, convert it: compare cost incurred against budget earned, not total budget. Budget earned is total budget times the physical percentage complete for that code.

• Budget $100,000, 50% of the work done, $48,000 spent — you are ahead.
• Budget $100,000, 50% of the work done, $68,000 spent — you are 36% over on the work performed, and the job is only halfway.

The second one projects to a $36,000 overrun at completion. The raw variance column would have shown $32,000 under budget. That is how a job can look fine and be failing.

The Second Question: Rate or Quantity?

Once you know a code is genuinely over, there are only two causes and they call for different responses.

Rate variance — you paid more per unit than planned. Material prices rose, you used a higher-cost crew, overtime crept in, a sub came in above the budget line. The work itself was as expected; the price was not.

Quantity variance — you used more units than planned. More hours for the same scope, more material for the same area, rework. This is the one that signals a real problem, because it usually means the estimate was wrong or execution is struggling.

Separating them takes unit data — hours and quantities, not just dollars. If your job cost system only carries dollars, you can see that something went wrong but never why, which is the main reason variance reports get ignored.

The Third Question: Is It Recoverable?

An overrun caused by scope the owner added is a change order you have not raised. An overrun caused by your own crew taking longer is a margin loss you absorb. Same number in the variance column, completely different conversation.

Ask on every material variance: did something change that the contract says we get paid for? Differing site conditions, owner-directed changes, design errors, delays caused by others. Contractors lose an enormous amount of recoverable cost simply by not asking the question while the evidence is fresh.

Which Variances Are Worth Your Time

On a job with fifty cost codes you cannot chase all fifty. Sort by dollars, not percentage. A 40% overrun on a $2,000 code is $800 and noise. A 6% overrun on a $400,000 subcontract is $24,000 and worth a meeting.

A workable rule: review anything where the projected overrun exceeds 1% of contract value, or $10,000, whichever is lower. Everything else gets looked at monthly in aggregate.

Where Variance Feeds the Rest of Your Reporting

A genuine variance is not just a job management issue — it changes your cost-to-complete, which changes your percentage complete, which changes revenue recognised on your WIP schedule.

A contractor who identifies a $50,000 overrun and does not update the budget is reporting profit they will not earn. That reverses at close-out as margin fade, and if a surety is watching your WIP schedule, repeated fade is the pattern that costs you bonding capacity.

Making It Routine

Monthly, per job, in a short conversation between the PM and accounting. Variances over threshold get a cause — rate or quantity — and an owner. Anything potentially recoverable gets a change order raised or a documented decision not to.

Fifteen minutes a job. It is the cheapest margin protection available to a contractor, and almost nobody does it consistently.

If your variance reports are not producing decisions, our Job Cost & WIP Diagnostic looks at why.