Changelog

What changed, and when

Written by the people who made the change. Fixes are listed with the same weight as features, because a fix is usually what somebody was waiting for.

Version 1.3

July 22, 2026

The weekly saves list, and export

Growth and Scale now get the weekly accounts-to-save list: the ranked handful worth a person's time this week, with each account's factors and drafted play one click away.
Added CSV export of scores, bands and factors, so the work leaves Churnora as cleanly as it arrived. Your data is yours, and now the door says so.
Scoring a 5,000-account book now finishes comfortably inside the import request. The scorer runs in-process, so there is no queue to wait on.

Version 1.2

June 9, 2026

Factors you can check

Every factor now shows the observable number behind it, like last sign-in 31 days ago or 2 failed payments in 90 days, so a claim about an account can be checked against the account.
When no language model is configured, explanations now fall back to code-written phrasing and the interface labels them as such. The score never depended on the language model, and now the words are honest about their source too.
Published scripts/train-model.ts, the exact script that fits the shipped model, alongside the method on How the model works.

Version 1.1

April 28, 2026

A free tier with the real model

The Free tier now scores up to 200 customers a month with the same model the paid tiers use. The limit is enforced server side, in code, and an oversized import explains itself rather than silently truncating.
Import got forgiving: header names are matched loosely, quoted fields with commas work, and every skipped row produces one plain-language reason.
Added the bundled sample book, 48 believable accounts, so you can feel the ranked list before exporting anything from your own systems.

Want something on this list? Tell us. Most of what ships here started with a customer-success team pointing out something the product got wrong about their week.