The estimate is a prediction nobody checks
Every quote you send is a prediction: this much labor, these parts, this price. Then the job happens — and almost nobody goes back to see whether the prediction was right. The tech finishes, the invoice goes out, and the operation moves on to the next job with the estimate that was quietly wrong still sitting in the price book, ready to be wrong again next time. That unchecked gap between what you quoted and what the job actually cost is one of the most expensive blind spots in field service, because it means the same job type loses money the same way, over and over, with no feedback loop to catch it.
This is a different question from which jobs make money. Job costing tells you whether a completed job cleared a margin. Estimate-versus-actual variance tells you whether your quote was any good — whether the four hours you priced became seven on site, whether the parts list you quoted matched the parts the tech actually burned. A job can clear a margin and still reveal a broken estimate; a job can lose money purely because the quote was wrong, not because the work was. You need both lenses, and most shops have neither.
Where the quote and the reality diverge
When you actually line up the estimate against what happened, the gaps cluster in a few predictable places:
- Labor hours. The most common and most expensive miss. You quoted three hours of labor; the job took five because of access problems, a stuck fitting, or a complication the estimate never anticipated. Two hours of unpriced labor, multiplied across every job of that type, is a steady margin bleed nobody sees because nobody compares.
- Parts. The estimate listed two parts; the tech used four once the panel came off and the real condition showed. The extra parts either got eaten or got captured as a change order — and if your estimate keeps under-listing parts for a given job type, that's a fixable pattern, not bad luck.
- The drive. A quote that ignores windshield time for a job across town underprices the real cost of showing up. If a job type is consistently far-flung, the estimate should carry the travel, not pretend the tech teleports.
None of these are random. They're patterns, and a pattern you can see is a pattern you can price into the next quote.
You can only compare what you captured honestly
The whole exercise depends on having a real "actual" to compare the estimate against — and that's exactly where most operations fall down. If the tech's hours are reconstructed from memory on Friday and the parts are a vague recollection, your "actual" is fiction and the comparison is worthless. Garbage actuals make every variance meaningless.
This is why honest labor capture is the foundation under variance analysis, not a separate concern. In Hosting Field, both the estimate and the job carry line items for labor, parts, and expenses with live totals, and the job's status timestamps (en route → on site → complete) record the real time the work took. So the estimate's quoted total and the job's actual billed total are both sitting there as structured numbers — the comparison is a matter of lining up two records that already exist, not excavating one of them from someone's memory. Be honest about the boundary: Hosting Field doesn't ship a canned variance report that flags the gap for you. It gives you the quoted line items and the actual line items as clean, captured data; putting them side by side and reading the pattern is your discipline. But that discipline is only possible because the actuals were captured truthfully in the first place — which is the part most shops skip and the reason their estimates never improve.
Close the loop, or keep paying the same tax
Variance is only worth measuring if it changes the next estimate. The whole point is the feedback loop: the job you just finished should make the next quote of its kind more accurate. A few disciplines turn the comparison into a fix:
- Re-price the job types that consistently run over. If your standard install quotes for four hours but the actuals cluster at six, the estimate is wrong — not occasionally, systematically. Move the price book number to reflect reality. You're not raising prices for fun; you're charging for work you're already doing for free.
- Separate "the estimate was wrong" from "the job genuinely grew." A job that ran long because the tech found a second problem isn't an estimating failure — it's scope that should have become a change order. Don't "fix" your estimate to absorb scope growth that should have been billed separately; that just hides the real lesson, which is to capture changes on site.
- Watch the estimator, not just the estimate. If one person's quotes are reliably tight and another's are reliably optimistic, that's a coaching gap, not a price-book gap. Variance data, sliced by who quoted, tells you which.
The trap to avoid: padding every estimate to cover the worst case. Quote everything for the seven-hour disaster and you'll lose the jobs that were genuinely four-hour jobs to a competitor who priced them honestly. The goal isn't a fatter buffer on everything — it's accurate estimates per job type, tightened by what the actuals taught you.
What to measure
- Average labor variance by job type — quoted hours versus actual hours, grouped by the kind of work. The biggest and most fixable miss; this is where you find the job types your price book is lying about.
- Parts variance — quoted parts cost versus what the job actually consumed. Consistent under-listing means your estimates need a more complete parts picture for that work, or your techs need to surface additions as change orders on site.
- Estimate-to-invoice delta — the gap between the quoted total and the final billed total, as a percentage. Small and centered on zero means your quoting is honest; consistently negative means you're systematically underpricing and feeding margin away one job at a time.
An estimate you never check is a mistake you're doomed to repeat. The job you just finished is a graded answer to a prediction you made — and reading that grade is how your quoting actually gets better instead of being wrong in the same comfortable ways forever. Capture the actuals honestly, line them up against the quote, and let what the jobs teach you tighten the next estimate. Do it for a quarter and the job types that quietly lose money will stop hiding, because the gap between what you said and what happened will finally be a number you can see.