// blog
Product analytics, written for people who have to decide
Event design, funnel diagnosis, retention maths and where a language model genuinely helps. Practice rather than product news — everything here is usable whatever tool you run.
// latest
Asking your data in plain English
A language model turns a question into a query. It cannot tell you whether the answer means anything — and that division of labour is the point.
Every post here follows the same rule we hold the product to: say what the number means, not just the number. Definitions, windows and denominators are stated, and the examples are ones you can run against your own data today.
// all posts
Everything we have published
Showing 4 more posts after the latest one.
What retention actually measures
N-day, bracket and unbounded retention answer three different questions. Pick the curve that fits your product, then judge where it flattens.
Your first activation metric is probably wrong
The milestone you chose in a meeting usually correlates with retention without causing any of it. How to find the behaviour that does.
How to read a funnel without fooling yourself
Every funnel hides five arbitrary decisions. Here is how to spot them, and how to rank steps by the users you could realistically recover.
Designing an event taxonomy that survives a year
A naming convention, a property contract and a five-minute review that keep your event list readable long after the people who wrote it move on.
Try the ideas on your own data
Every post here is written against the analysis Alitycs does by default. Send an event and check the maths yourself.
Written by the team that builds the product. No guest posts, no sponsored placements.