Why do door schedules and plan counts disagree?
Because the schedule and the plans are edited separately, under deadline, by different hands. A revision lands mid-bid and updates the sheets but not the schedule. A row gets added that nobody ever places as a symbol. A symbol gets copied with a duplicated unit plan and never gets a row. The legend keeps types that drifted from both. Estimators we work with spend 15 to 20 hours per job at roughly 2,500 openings, and hand reconciliation at that scale is sampling, not checking. The reliable fix is counting all three sources independently and reconciling the differences.
Where do the mismatches come from?
A door schedule and a floor plan describe the same building, but nobody edits them as one document. Under a bid deadline, four failure modes pull them apart.
- Revisions landing mid-bid. An addendum touches the plan sheets, the schedule ships a version behind, and every count finished last week is now against the wrong set.
- Schedule rows nobody placed. A door gets added to the table, the symbol never gets drawn, and the row rides along looking authoritative.
- Symbols nobody scheduled. A unit plan gets copied across a floor, the doors copy with it, and the schedule never hears about them.
- Legend drift. The legend defines types that were renamed, resized, or dropped, so the same mark means different things on different sheets.
Why does hand reconciliation fail at 2,500 openings?
Estimators we work with spend 15 to 20 hours per job at roughly 2,500 openings, and they report 70 percent of that time going to data entry. There is no budget left to check every schedule row against every symbol on every sheet, so reconciliation becomes sampling: spot-check the floors that look suspicious, trust the rest.
Sampling has a failure mode. The mismatches cluster exactly where nobody looks: the copied unit plan, the addendum sheet, the type nobody uses anymore. And when a revision lands mid-bid, the reconciliation already done is stale, with no honest way to know which counts survived.
How does independent triple-counting surface every mismatch?
Count the three sources separately, then compare. Quoting.ai Takeoff parses the schedule as a table, detects and tallies plan symbols sheet by sheet, and indexes the types the legend defines. The 4 Eyes check reconciles the three counts and turns every disagreement into a line item with sheet references: this row has no symbol, this symbol has no row, this type's counts do not match.
Openings carry apartment-based numbers, 1A01 rather than a random ID, so a flagged mismatch points to a location a submittal can reference. The estimator works the discrepancy list instead of re-counting the set, and when an addendum drops, the revised set runs again in the browser and produces a fresh list.
What should you do when the counts disagree?
Do not average, and do not take the bigger number to be safe. Each mismatch is a specific question with a specific answer: the row is real and the symbol is missing, the symbol is real and the row is missing, or a revision made one of them obsolete. Resolve what the set itself answers, and send the rest to the GC or architect as questions before bid, with sheet references attached.
Whatever stays unresolved becomes a written qualification on the bid. That converts a silent gap somebody eats after contract into a priced, visible decision.
Related questions
Which number is right, the schedule or the plans?
Neither by default. Each is edited separately, so each carries its own errors. The disagreement is the finding: resolve it per opening, and put anything the set cannot answer in front of the GC or architect before bid.
Does the 4 Eyes check run on windows too?
Yes. Doors and windows are the live trades in Quoting.ai Takeoff. Walls and framing are in development, and early access is open.
What happens when an addendum lands mid-bid?
Upload the revised set and the three counts run again. Review becomes the new discrepancy list, not a second hand count of the whole set.
See it on your own work
Distributors: two steps and a kickoff call. Estimators: upload a plan on a live trade. Either way, the product proves it or it does not.