AI quantity audits: ratio checks that catch bad numbers

The ratio and consistency checks an automated quantity audit runs on a sitework takeoff, and how to read the results without chasing false positives.

5 min read

Key takeaways

  • Ratios between related quantities are the fastest error detector.
  • Pipe per structure, base per paved area, and curb per lot area are the workhorses.
  • A flagged ratio is a question about scope, not proof of an error.
  • Historical project ratios beat generic industry ranges.

Quantities on a sitework project are not independent. Base stone follows paved area. Structures follow pipe footage. Curb follows the paving perimeter. When one of those relationships is far outside its normal range, something is usually missing — and the ratio finds it in seconds where line-by-line review takes an hour.

The core ratios

RatioTypical rangeWhat an outlier suggests
LF pipe per structure200–400 LFMissing structures, or missing runs
Base stone CY per paved SY≈ SY × in ÷ 36Wrong thickness, missing section
Curb LF per paved SF0.02–0.05 LF/SFMissing islands or perimeter curb
Silt fence LF vs disturbed perimeter≈ 1.0–1.3×Incomplete perimeter control
Topsoil respread SF vs disturbed − paved≈ 1.0×Deduction error
Inlets per acre paved3–8Missed drainage structures
Earthwork CY per acreProject-specificCompare to similar past jobs

These ranges are starting points, not standards. A tight urban infill site and a rural detention basin project have legitimately different ratios. The value comes from noticing the outlier and being able to explain it.

Reading a flagged ratio

  1. Ask what would make the ratio true — usually a specific missing scope.
  2. Check the sheets for that scope before adjusting anything.
  3. If the project genuinely differs, record why and dismiss the flag.
  4. If it is an omission, measure it and re-run the audit.

Build your own baselines

Generic ranges catch gross errors. Your own history catches subtle ones. Once you have measured a dozen similar projects with a consistent condition library, your ratios become the benchmark: this developer's retail pads always run about 90 CY of earthwork per 1,000 SF of building, and a bid at 40 is worth a second look.

  • Keep condition names stable so ratios are comparable across jobs.
  • Record project type and size alongside the quantities.
  • Compare new bids against the two or three most similar past projects.
  • Update baselines as your typical work changes.

What ratios cannot do

A ratio check cannot tell you a quantity is 8% low; it only catches order-of-magnitude and scope-level problems. Precision still comes from calibration and careful tracing. What the audit buys you is confidence that nothing large is missing — which is the failure mode that actually loses money.

Doing this work in TakeoffAI? AI review console.

Ready to run your next takeoff?

Create an account and turn your first plan set into measured quantities today.