September 19, 2026
Enrichment questions become probes you write — one input, one lookup, and as many questions as you want over it, so five questions about a store cost barely more than one; evidence and confidence are now per answer, so a weak answer no longer costs the good ones beside it; the profile page shows what each value was read from and why a missing one is missing; and editing a probe's wording now correctly re-asks it.
Ask a probe five things instead of writing five probes
AdminSettings → Data → Enrichment used to have two sections: a catalogue of ready-made probes, and a list of one-off questions. A shipped probe like Site read answered nine questions from a single page read — but a question you wrote yourself could only ever produce one property, so asking five things about a store meant five probes, five reads of the same page, and five charges.
Now there's one list, and everything in it is a probe. Add probe gives you what Site read already was: one input, one lookup, and a list of questions answered from it.
| Question | Answer type | Property |
|---|---|---|
| What platform is this store built on? | A few words | platform |
| Roughly how many products do they list? | A number | product_count |
| How many reviews does their best seller have? | A number | review_count |
| Do they sell wholesale as well as retail? | Yes or no | sells_wholesale |
One page read, one charge, four properties — and one request at the store's server rather than four. Because the questions share a lookup, adding one to a probe you already run is close to free. If you're wondering whether something is worth asking, put it on an existing probe rather than standing up a new one.
Everything you'd already written keeps working: a question you saved yesterday is now a probe that asks one question, and you can add more to it.
Evidence and confidence are per answer
AdminThe guardrails that make an answer safe to filter and score on — a declared answer type, required evidence, a confidence floor — now apply to each question rather than to the lookup as a whole. One question the pages don't support no longer costs you the four they do.
A profile's Looked up card shows this directly: each value lists the pages that answer came from, and each question the lookup couldn't answer is listed with the reason — nothing cited, confidence too low, or not stated on the pages read. A blank used to be a blank; now it tells you which kind it is.
The five-profile preview shows the same thing before you save: every question, answered or dropped, for each of the five.
Editing a probe's questions now re-asks them
Admin FixedEditing the wording of a question you'd already saved left the stored answers in place until the refresh window expired — up to sixty days of a property whose meaning no longer matched the question that produced it, still being read by your rules and scores.
Changing what a probe reads, how it reads it, or what it asks now correctly throws out its stored answers and re-runs it. Renaming a probe, pausing it, or narrowing its scope still costs nothing, because none of those changes what a past answer meant.
Coverage works for probes you wrote
AdminThe coverage line — "runs on 8,200 of 12,000 profiles (68%)", the number that tells you which signup question is worth adding — was only shown for the ready-made probes. It now appears on every probe, including your own. So does the property picker's ranking, which now puts likely properties first when you point your own probe at something.
September 21, 2026
A usage metric now measures the step that multiplies rows and treats the steps after it as lookups, so you can sum a charge's amount while filtering on its payment intent — and the columns that would have given you a wrong total are no longer offered.
September 18, 2026
Enrichment looks up what the outside world says about an account — whether their site is live, what it's built on, how much they sell — and stores it as ordinary properties your rules and scores already understand, with a coverage line that tells you which signup question would unlock the most and a monthly ceiling that pauses rather than surprises; monitors land as the engine behind "something is wrong right now" — three of them start watching your accounts in observe mode — funnels can enter on one, funnels say why an enrollment left and can time one out, the Activation and Reactivation stage pages show a stage conversion rate, Intent is now Momentum with one momentum model per funnel edited from the funnel, and the inherited commerce scaffolding is gone — the Catalog, Orders, Customers, Discounts, Storefronts, Inventory routing, Point of sale and Analytics entries leave the sidebar, the products API and its SDK and MCP tools are removed, the Overview shows where your accounts sit in their lifecycle, ⌘K reaches every lifecycle stage and settings page, and webhooks subscribe to `lifecycle.evaluated`.