The anatomy of a double count

On a recent school tender, one CAV box appeared on two drawings: once on the ventilation schematic, once on a plan. The manual estimate priced it twice. Nothing about the mistake was careless; it is the natural output of the way tender packs are put together and the way humans read them.

Why the documents overlap

A drawing set is deliberately redundant. The schematic shows the system logic, the plans show locations, sections show what the plans cannot, and the specification restates in prose what the schedules tabulate. The same physical box legitimately appears in four places. The estimator's job is to recognise that four appearances are one asset, and they do it with a tag register, a highlighter and concentration.

Concentration is the weak link. Tender deadlines put this work at the end of long days. Two drawings are read on Tuesday and Thursday. The tag is transcribed slightly differently in two rows of the spreadsheet. Revision C added a plan that duplicates part of the schematic. None of these require incompetence, just volume and time pressure.

Why it matters more than it looks

One duplicated CAV box is a few hundred pounds of unnecessary price. The pattern is what costs: if the same overlap duplicates a run of terminal units, a floor of fan coils, or the metering, the error scales with the equipment count. It cuts both ways. A double count inflates your price and loses the job; the mirror-image error, treating two genuinely different boxes as one, wins the job with kit missing from the number. Both errors survive review easily, because a points schedule with 900 rows does not invite line-by-line re-reading.

How software catches it

Counting every tag once across every document is a bookkeeping problem, and bookkeeping is what software is for. Tenderfly reads the whole pack and gives each piece of equipment one identity, however many documents it appears in. Two appearances of the same tag merge into one asset with two pieces of evidence; two similar tags that are genuinely different stay separate. When the count disagrees with a human one, the disagreement is inspectable: the estimator on that school tender asked why the numbers differed, and the answer was the drawing, opened at the page, with the tag highlighted.

That last part is the point. A different number is only useful if you can find out why it is different in seconds rather than by re-reading the pack. Double counts stop being embarrassing discoveries after submission and become one-click checks before it.