The short version
- Scale is still the most expensive single error, and it is invisible because everything stays internally consistent.
- Check the drawing index, not the folder. Missing sheets do not announce themselves.
- Overlap between trades double-counts or drops work at the boundary, and the boundary is where the money is.
- Anything measured to a round number is worth a second look. Real buildings rarely produce round numbers.
1. The scale is wrong and everything agrees with it
A drawing calibrated off a printed scale bar on a PDF that was reduced to fit A3 gives you a takeoff that is internally perfect and uniformly wrong. Every ratio checks out. Every subtotal cross-foots. The building is just eleven percent smaller than it is.
The fix is not to trust the scale bar. Calibrate on a dimension string, ideally a long one, then verify against a second dimension on a different part of the sheet. If the two disagree, the sheet has been stretched non-uniformly and no calibration will save it. Go back and ask for the original.
2. Sheets that were never in the set
You were sent seventeen drawings and measured seventeen drawings. The index lists twenty-two. The three you are missing are usually the enlarged plans, and enlarged plans exist precisely because that part of the building is complicated.
Read the index first, tick sheets off against it, and put anything missing in writing before you start rather than after. It is a two minute check that changes who owns the gap.
3. Revision clouds treated as decoration
A set issued at Revision C where you measured Revision A is a known problem and people watch for it. The subtler version is a set where different sheets sit at different revisions, which is extremely common. The architectural plan says Rev C, the structural plan under it says Rev A, and the column that moved on one has not moved on the other.
Record the revision of every sheet you measure, on the sheet list, in the workbook. When the addendum lands you then know exactly which measurements are affected instead of re-measuring everything in a panic.
4. Double counting at the trade boundary
The concrete package includes the plinth. So does the masonry package, because the mason priced from the elevation. Or neither includes it, because each assumed the other had. Boundaries between packages are where quantities get both duplicated and dropped, and it is the same root cause: nobody wrote down where one scope stops.
- Slab edge thickening: structural concrete or groundworks?
- Builders work in connection with services: the service trade or the main contract?
- Fire stopping at a service penetration: the service trade, the drylining, or a specialist?
- The screed: floor finishes or the concrete frame?
- Backfill against a retaining wall: earthwork or the concrete package?
Each of those has a defensible answer and none of them has a universal one. Write the answer down on the estimate, in an exclusions list, and the boundary stops being expensive.
5. Work that is real but not drawn
Temporary works, access, protection to finished surfaces, out-of-hours restrictions, and the crane time to get a plant room unit onto the roof. None of it appears on a drawing, all of it appears on an invoice. The specification and the preliminaries usually carry it, and neither is where a takeoff naturally looks.
6. Areas measured to the wrong face
Gross internal, net internal and structural gross differ by the thickness of the walls, which sounds trivial until you multiply it across forty floors. Floor finishes measured to the structural grid rather than the finished face will over-order by roughly the perimeter times the wall thickness, every time.
Pick a convention, state it at the top of the takeoff, and use the same one for the finishes as for the substrate under them.
7. Round numbers
A line reading exactly 500 linear feet is either a coincidence or a placeholder somebody meant to come back to. Real geometry produces 487 and 512. We search every takeoff we review for suspiciously round quantities before we look at anything else, and it finds something more often than it should.
8. A count that measured symbols, not devices
Automatic symbol counting is fast and it is right most of the time. The failures are specific: one symbol used for two different devices at different sizes, a legend that changes between sheets, ganged devices drawn as a single symbol, and devices on a demolition sheet counted as new. A count that comes back as a clean multiple of the room count deserves suspicion.
9. The estimate was checked, but only against itself
Cross-footing a spreadsheet proves the arithmetic, not the measurement. The check that actually finds things is a ratio check against a comparable job: cost per square metre, rebar tonnes per cubic metre of concrete, board area per linear metre of wall, devices per square metre. If any of those sits a long way off a job you have built, something is wrong and the spreadsheet will never tell you.
A takeoff that agrees with itself is not verified. It is only consistent.
What a review looks like
- Reconcile the sheet list against the index and record every revision.
- Re-calibrate two independent dimensions on each sheet used for measurement.
- Run the ratio checks against a benchmark job before opening any line detail.
- Search for round quantities, zero quantities and quantities with no unit.
- Read the exclusions against the specification, looking for work that sits in neither.
That sequence finds most of the nine above in a couple of hours, which is a good trade against finding them after the bid has gone in.
Written by the estimating team at QuantX Estimation. Since 2016 we have measured and priced construction work for contractors in the United States and United Kingdom, at a fixed fee agreed before we start.
