Three thousand bid files, priced by one person.

Sadoff is a telecom parts dealer. Bid files arrive in volume and every one has to be priced, and the pricing depends on judgment that lives with one person. Peter spent years learning which parts are worth what, and why, under conditions that change constantly.

Three thousand files. One expert. A business whose throughput was capped by how fast one person could read.

Speed was the visible problem. The real one was that the logic holding the business together existed nowhere except inside somebody's head.

We proved it before we priced it

The work started with a paid proof of concept, run over 27 development days. It was not a demonstration. It had to establish whether the hard parts were possible at all: parsing bid files of inconsistent quality, matching parts against supplier data that rarely agrees with itself, reasoning about price across multiple agents rather than one model, pulling live market intelligence from eBay, BrokerBin and Alibaba, and holding on to what it learned between runs.

By the end there were 37 commits and 24 completed stories. More usefully, there was a measured build velocity.

Estimates from evidence, not instinct

Most software estimates are informed guesses. This one was arithmetic. The proof of concept showed how fast the work actually moved, so the production build could be sized at 97 days across seven areas of work, phased through stability, data foundation, pricing intelligence, then workflow and polish.

One senior architect was named in the proposal before any price appeared. Peter was in the room for every pricing decision, because the goal was to encode his judgment rather than replace it with something generic.

The same scope was offered three ways, differing only in how much was paid up front against how much was paid monthly. After launch, a fractional architect stayed available for changes inside the original scope.

How the work was shaped

If this sounds familiar

When the most valuable judgment in your business lives with one person, you have two problems at once. Everything waits for them, and the day they leave it goes with them. Encoding that judgment is a real build, and a short paid proof of concept will tell you what the real build costs before you commit to it.

How other engagements were run.