Amplitude

Promptfree, for everyone, and the only version there is

Copy the prompt for Claude CodeCopy the prompt for CodexCopy the prompt for Cursor

Copy

Do not rebuild this at company scale — but notice what you are actually paying for, because it is not the charts.

It is **shared definitions**: product, growth and marketing all meaning the same thing by "activated user", computed from one event stream, so two teams cannot arrive at a meeting with different numbers. That agreement is organisational, not technical, and it is the expensive half.

WHEN TO KEEP PAYING
- Several teams using the same numbers to make decisions
- Cohorts pushed into advertising and email tools continuously
- Hundreds of millions of events, where recomputing a funnel in seconds is genuinely hard

WHAT YOU CAN BUILD, AND SHOULD, WHATEVER YOU DECIDE

**A metrics definition layer**, which costs a week and is worth more than the subscription:
- Every metric defined once in a file in a repository: its name, its SQL, its unit, its owner, and a sentence saying what it means
- Reviewed like code, versioned, with a changelog. When "active user" changes, the history of that change is visible and dated
- Generated documentation listing every metric, its definition and its owner
- A test per metric that runs against a known dataset with a known answer, so a definition cannot silently drift

Then the reports are ordinary SQL over your warehouse:
- Funnels as a window function over ordered events, with strict or loose ordering stated
- Retention as a triangle, with the definition named because there are several and they disagree
- Cohorts as saved queries, computed at query time, never stored

THE CHECK NOBODY RUNS
Recompute last quarter's headline number with today's definition. If it differs from what was reported at the time and nobody knows why, you do not have an analytics problem, you have a definitions problem — and no tool fixes it.

THE ONE-LINE VERSION
Buy the platform if you need the scale. Build the definition layer either way, because that is the part that makes the numbers trustworthy.