WIREPITCH

Changelog

What shipped, and when.

Dated so you can see whether this thing is actually being worked on.

Reading the software a business runs

Scoring could only see how well-maintained a website was, so every market ended up scored on roughly the same thing. It now reads which third-party services a business has embedded in its own site — booking and scheduling, payments and giving, video and livestreaming, email marketing, operations software — which is a far better signal of how a business actually operates. "Already takes payments online" and "runs a specific scheduling tool" became things a scoring model can ask for.

Scoring models can no longer ask for something impossible

A model could propose a heavily-weighted signal matching a value the crawler can never produce — it would have read sensibly on the page and scored zero on every row. The signal catalogue is now generated from the extractors themselves, so a model can only ask for values that actually exist, and anything outside them is rejected and rebuilt.

Scoring is deterministic, and shows its arithmetic

Prospects used to be scored by a language model reading each site. They are now scored by a weighted model applied to stored signals. A score became a sum you can dispute line by line, it gives the same answer every time, and a prospect's own website can no longer argue for its own score. Every delivered row now carries which signals it hit and which it missed.

You see the scoring model before you pay

Send a brief and the site returns the model it would score your market with — named signals, weights, the threshold, and an explicit list of anything you asked for that we cannot honestly check. Free, and before any money is discussed.

Pricing rebuilt around scope

The old tiers were priced by pipeline stage, which meant nothing to a buyer. Pricing is now one ladder — a free scoring model, a $250 sample, a city, a state, a whole vertical — where the only thing that changes is how much of your market gets scored. Every rung includes the model.

Claims the site is no longer allowed to make

Removed a headline that wasn't defensible, a promise that every address was verified when verification was switched off, an unbounded refund promise, and a claim that a person reviewed every stage of every run. There is now an automated check that stops any of them reappearing.