TAKES
What should we track?
Every event is a question someone meant to ask.
EVERYONE'S RIGHT
Every answer
AlexThe Architect“One typed tracking plan. Every event generated from the schema.”
JoThe Cowboy“Nothing yet. Ship it, then add the event we wish we had.”
RileyThe Minimalist“The orders table. It already knows who paid.”
TaylorThe AI Native“Everything. The agent will find the question later.”
MorganThe Threat Modeler“Only what we'd be comfortable reading aloud to a regulator.”
ChrisThe Sovereign“Self-hosted analytics, our server. No third-party script on checkout.”
PatThe Enterprise Adult“Whatever the board deck asks for, from a vendor with a DPA.”
MayaThe Visionary“Every click. You can't A/B test what you didn't log.”
NoraThe User Advocate“Watch five sessions first. Then track what confused them.”
HugoThe Perfectionist“Fewer events, each with an owner, a schema and a test.”
RuthThe Maintainer“We have 412 events. Which ones did anyone open this year?”
GregThe Hedger“Track everything now. You can't backfill a question later.”
🎯 SUPER REASONABLE
Super Reasonable, the advisor who never takes a side
Tracking is cheap to add and expensive to trust. Start from the decisions the numbers will drive, check every number that matters against a source that cannot drift, like orders or payments, and watch a handful of real users before acting on a funnel nobody understands. Collect generously only where it is anonymous and someone will actually look. For everything else, the best event is one you never have to explain to a regulator.
- Which decision will this number drive? If none, don't track it yet.
- Does it match the money? Reconcile every funnel step that matters against the orders or payments table.
- Does anyone read it? Delete events nobody queried in 90 days.
- Could you defend it to a regulator? No purpose, no consent, no event.
📚 Seen in the field: Try everything · Nobody can find it · Resolved
🔬 IN THE FIELD GUIDE
