Work record / Stories

Catching my own five-times overcount

To prove the rebuilt estimating engine, I re-ran a project we had already priced the old way and found our own error.

After I overhauled the estimating engine to measure straight off a drawing's vector geometry, replacing a desktop approximation, I needed real proof it worked. I chose a 36,000-square-foot institutional campus that the team had already estimated once, with the old method. I ran the new system on the same project we had priced months earlier, not on something fresh and unproven.

The most credible proof that a better process works is catching your own earlier mistake. Producing a plausible number on a new job proves little. Putting the new method against our own prior estimate, on identical drawings, was the honest test.

On partitions, the new system came in at 1,834 linear feet. The earlier desktop estimate had 9,470. That is a five-times overcount, and it would have mispriced the bid. An independent blind pass confirmed the corrected figure to within about 2.4 percent.

The strongest evidence a new process works is watching it catch a real error your old one made. Plenty of people would rather bury the old number quietly than use it as the proof.

Story details

All stories · Work record