How a count is found, and how much to trust it

A count goes onto an order, so “the software found them” isn’t an answer anyone will accept in six months. Every automatic count therefore carries a record of how it was found, what was searched for, where, when, how many were first found, and what you did about them.

From most to least certain

Measured by hand

You clicked each one. As good as the drawing you read.

Drawing text

Found by reading the drawing's own words, such as the tag RL01. Exact matches, not guesses, but text can also appear in a legend or a schedule, so look at the pages.

Vector match

Found by matching the drawing's own lines against an example. Deterministic, but a judgement about shape: it counts things drawn the same way and the same size, and you can allow close ones, which are then shown to you.

Visual match

Found by comparing pictures (for a scan). Inferred, and every match is shown for you to check.

What every automatic count promises

• It’s shown to you before it’s made, so what’s counted is what you saw.
• It’s an ordinary count: you can move it, delete marks, add to it and put it in a take-off group.
• It records how it was found, including how alike you allowed, any areas you left out, and how many you ticked off.
• A sheet it couldn’t look at is named, not silently counted as zero.
• Matches you ticked off are recorded as rejected by you; sheets you left out or areas are recorded as scope. Rejections aren’t credited to you unless you made them.