What Scope Creep Actually Costs and Where It Starts

Why this matters

Every shop can name the hours it gave away last quarter, and almost none can name what those hours displaced. That second number is usually the bigger one, and it is invisible because it lands on a different job than the one that caused it. The other half of the problem is that shops look for the cause in the wrong place. The famous culprit, the customer who says "while you're here," turns out to be a minor contributor. The dominant source is written in your office, before anyone drives anywhere, in a line item that names a task without naming the condition it assumes.

A sibling article covers the estimate as a boundary and the language that erodes it. This one is about the arithmetic and the origin trace, which is a different question with a different answer.

The first cost, which everyone measures

A shop with three field crews pulled a full quarter: 62 completed jobs, 1,240 quoted hours, 1,388 actual hours. That is an overrun of 148 hours, or 11.9% over quote.

They split the 148 by whether it was recovered: 34 hours went out as priced change orders, and 114 hours were absorbed. The absorbed figure is 9.2% of quoted hours, and it is the number that shows up in a job-costing review.

Stop there and the conclusion is that scope creep cost them 114 hours. That conclusion is wrong by a large fraction, and the reason has nothing to do with the arithmetic.

The second cost, which almost nobody measures

The shop runs a board with committed start dates. An hour added to a job in progress does not evaporate; it pushes the crew's next start. Over the same quarter they recorded 19 rescheduled starts caused by an in-progress job running past its window:

• 11 were absorbed by the next customer without any cost to the shop.
• 5 produced a wasted mobilization: the crew traveled, could not start, and returned another day. At 1.8 hours of travel and setup per wasted trip, that is 9.0 crew hours that produced nothing at all.
• 3 were cancelled outright by customers who hired somebody else. Those 3 jobs represented 46 quoted hours that left the board, of which the shop refilled 18 hours at short notice, leaving 28 quoted hours unsold.

Eleven plus five plus three is 19.

Now be careful about what can be added to what, because there are three different currencies here and adding them all up would be the easiest mistake in this article.

One honest total: 114 absorbed hours plus 9.0 wasted mobilization hours is 123 crew hours that were paid for and produced nothing billable. Same currency, straight sum.

The 28 unsold quoted hours are not in that currency. They are gross sold work the shop no longer gets to do. Losing them costs the margin on 28 hours, not 28 hours, and only for as long as the slot stays empty. Reporting them as though they were equivalent to absorbed labour would overstate the finding by a wide margin.

Even stated carefully, the shape holds: the visible cost was 114 hours and the fuller picture is 123 crew hours plus 28 hours of sold work lost plus three customers gone. The displacement is not a rounding item sitting beside the direct cost. It is roughly the same order of magnitude, and it lands on jobs that look, in the job-costing report, like they went fine.

Where it actually starts

The shop then did the part most never do: they traced each of the 114 absorbed hours back to the first moment the extra work appeared, not to the moment it was noticed.

Seventy-one plus twenty-nine plus fourteen is 114, and the shares are 62%, 25% and 12%, which sum to 99% before rounding.

The customer additions everyone blames are 12% of it. The largest bucket by a distance is a line written in the office that says something like replace the shutoff, repair the damaged section, or service the unit, with no statement of what condition that price assumes. When the tech arrives, the condition determines the work, and there is no boundary anywhere in the document to compare it against. The tech is not creeping. He is doing the only thing the line can mean given what he found, and nobody wrote down what it was supposed to mean given something else.

That is the origin claim: creep starts where a scope names a thing without naming its state. It is upstream of the site, upstream of the customer, and it is entirely within your control.

The 25% bucket is different in kind and should not be lumped in. A concealed condition discovered mid-job is not a scoping failure, it is an unknown that either had a disposition before the price or did not. When an opening exposes suspect thermal system insulation, work in that area stops and the material is left undisturbed to be assessed and handled under 29 CFR 1926.1101 rather than cut through carefully; when it exposes an unexpected energized conductor, the circuit is de-energized, locked and tagged under 29 CFR 1910.333(b)(2) and proved dead with a meter checked live on a known source before anyone continues. Neither of those is a change-order conversation until the hazard is handled.

What they changed

One thing, deliberately, so the next quarter would mean something: they rewrote their 20 most-used catalog lines to carry a condition assumption and a named alternative. Nothing about the change-order process changed. Nothing about crew scheduling changed.

The shape of a rewritten line:

"Replace fixture shutoff. Price assumes the existing supply is sound and the stop is accessible without opening finished surfaces. A corroded supply or a concealed stop is additional at the stated rate, quoted on site before proceeding."

Two sentences. The second one is the entire fix, and it does two jobs at once: it gives the tech a boundary to recognise, and it gives the customer a warning before they are standing in a torn-open bathroom hearing a number for the first time.

The next quarter

1,310 quoted hours. Additions totalled 113 hours: 52 priced and recovered, 61 absorbed.

Compare on the same basis. Quoted volume differed by about 6% between the quarters, so use percentages of quoted hours rather than raw counts, which normalises that difference:

Read the recovery row, not just the absorbed row. Recovery went from 34 of 148 to 52 of 113, from 23% to 46%, roughly double. That is the mechanism working as designed: the condition language did not make extra work disappear, it made extra work visible early enough to price. The absorbed figure fell partly because additions fell and partly because a larger share of what remained got billed.

Total additions also fell, from 11.9% to 8.6% of quoted hours, and the honest explanation is that writing a condition assumption into a line forces someone to think about the condition at quote time, which surfaces some of it before the truck rolls.

One quarter at one shop is not proof, and the two quarters were different seasons with a different job mix. What makes this comparison better than most is that they changed exactly one thing and left the process that would most obviously confound it alone.

What changes the answer

A shop with no committed start dates. If your board is loose and customers accept moving windows, the displacement cost above mostly does not apply to you, and the absorbed hours really are close to the whole cost. That is a genuine structural difference, not a reason to skip the origin trace.

Time-and-materials work. Creep on T&M is not a cost, it is revenue, and the risk moves to a different place entirely: the customer's tolerance for an invoice they did not expect. The origin fix still helps, because the condition sentence is what keeps that invoice from being a surprise.

Long-duration project work with a formal change process. There the recovery rate is the number to watch rather than the absorbed hours, and a low recovery rate usually means the change process is slower than the crew, so work gets done before paperwork can catch it.

A single-crew shop. Displacement is more concentrated, not smaller: one crew running long delays exactly one next start, but you have no second crew to absorb it, so the 11 quietly-absorbed reschedules in the case above would not have been quietly absorbed.

How to verify you got this right

The audit that produced the finding is worth running yourself, and it is not the same as a job-cost review.

• Trace to first appearance, not to discovery. For each absorbed block of hours, ask what document or moment first allowed the ambiguity, then classify it. A tech noticing a corroded supply is a discovery; the line that did not say what supply condition was assumed is the first appearance. Reviews that classify by discovery put almost everything in the concealed-condition bucket and conclude nothing is fixable.
• Count rescheduled starts as a separate metric. If nobody records them, the displacement cost does not exist in your data and you will keep sizing the problem at the absorbed number.
• Check your recovery share, not just your overrun. A shop absorbing 9% and recovering 20% has a different problem from one absorbing 9% and recovering 60%.
• Pull your ten most-used catalog lines and read them as a stranger would. For each, ask what condition the price assumes and whether that assumption is written anywhere. If you have to explain it out loud to answer, it is not written.
• Re-run the trace one quarter after you change the lines. The bucket shares moving is the evidence that the fix hit the actual origin rather than just tightening the paperwork downstream of it.

References

• 29 CFR 1926.1101, asbestos (construction), including presumed asbestos-containing thermal system insulation; 29 CFR 1910.333(b)(2), work practices for electrical work
• See related: Scope Creep Starts at the Estimate
• See related: How to Spot Scope Creep in Your Cost Data
• See related: The Catalog Entry That Invites Scope Creep