Skip to content
Quoting.ai

How do you count doors from architectural plans automatically?

Automatic door counting reads the same three sources an estimator reads. The door schedule is parsed as a table, the door symbols are detected on every plan sheet, and the legend is indexed as the set's type dictionary. The three counts are then reconciled, and every disagreement becomes a listed discrepancy for a human decision. Quoting.ai Takeoff runs this in the browser on the full drawing set and numbers each opening by apartment (1A01, 1A02) so the count carries straight into submittals. The manual baseline: estimators we work with report 15 to 20 hours per job at roughly 2,500 openings, 70 percent of it data entry.

How long does counting doors by hand take?

Estimators we work with report 15 to 20 hours per job on sets with roughly 2,500 openings: tagging symbols sheet by sheet in Bluebeam, then retyping what the door schedule already says. They put 70 percent of that estimating time on data entry, not judgment.

The hours are only half the cost. Nobody re-verifies thousands of hand tags against the schedule at the end of a count that long, so the count ships spot-checked at best. The doors that get missed are the ones that exist in only one of the set's three sources.

How does automatic door counting work?

Four passes over the set: three independent counts and one check. The output is a count where every door ties to a sheet, a type, and a schedule row, plus a short discrepancy list instead of an act of faith.

  • Schedule parsing: the door schedule is read as a structured table, every mark with its type, size, and hardware set.
  • Symbol detection: door symbols are found and tallied on every plan sheet, each one tied to its sheet and location.
  • Legend indexing: the legend's defined door types become the set's dictionary, so every detected symbol resolves to a type instead of a guess.
  • Reconciliation: the three counts are cross-checked, the 4 Eyes check. A schedule row with no symbol, a symbol with no row, a type with mismatched counts: each becomes a listed discrepancy with sheet references for the estimator to resolve.

Why upload the full set instead of cropped sheets?

Because the three sources never live on the same sheet. The schedule sits on its own page, the symbols spread across the unit plans, and the legend is wherever the architect put it. Crop the set down to floor plans and the cross-check is gone: the tool can only count what it can see, which is the same failure mode as hand counting.

Quoting.ai Takeoff runs in the browser on WebGPU and reads the full set as uploaded. No desktop install, no sheet-by-sheet prep, no deciding in advance which pages matter. Upload the set; the takeoff starts.

How do counted doors get numbered for submittals?

A door numbered 847 is a dead end the day the GC asks which door that is. Quoting.ai Takeoff numbers openings by apartment, 1A01 for door 01 in unit 1A, so the takeoff, the submittal, and the field all use the same address.

Random IDs generated at count time fail the moment the count leaves the estimating desk. Apartment-based numbering means the door list you bid from is the door list you submit, and the one the field checks openings against.

Related questions

Which trades does automatic counting cover today?

Doors and windows are live. Walls and framing are in development, with early access open. Trades ship when they handle dense real-world sets, not before.

Do I need to install anything to run a takeoff?

No. Quoting.ai Takeoff runs in the browser on WebGPU. Upload a plan set and it starts working.

What happens when the schedule and the plans disagree?

The disagreement is listed, never averaged away. Every schedule row without a symbol, symbol without a row, and legend mismatch is surfaced with sheet references for the estimator to decide. That is the 4 Eyes check.

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.