Field note · 7 min read

How to build an event taxonomy that survives product change

A useful tracking plan is a decision system, not a long inventory of clicks. Here is how to keep yours legible as the app evolves.

A calm modern workspace with notebooks and monitors

Useful app analytics begins with a precise question and a shared interpretation of evidence. The following approach is designed to be applied by a working product team, then adapted to its market, customers, and data maturity.

Start with decisions, not screens

Screen-by-screen tracking reproduces the interface in data and becomes obsolete whenever navigation changes. Begin instead with recurring decisions: where activation fails, which behaviours predict durable value, and what product change deserves investment. Each event should support at least one named decision or dependable operational check.

Use a compact grammar

Choose stable object-action names such as project_created or report_exported. Define required properties centrally, separate user properties from event context, and document allowed values. A compact grammar helps engineers predict names and gives analysts fewer exceptions to memorise.

Treat change as normal

Give every event an owner, status, first-seen date, and change note. Deprecate rather than silently repurpose an event. A short monthly review of unknown properties, sudden volume changes, and unused events prevents gradual decay while keeping governance proportionate.

A final working habit

Write the intended decision beside every analysis. Record the data limits, alternate explanations, and what new evidence would change the recommendation. This small practice makes review more honest and preserves context for the next person who encounters the question.

← Return to all insights

Apply it to your app

Need a clearer answer from your product data?

Bring us the question