Harriet

Experimentation analyst · Job role: CRO specialist

Tells you if your test really won, with every sum shown and a clear call.

I read the results of your A/B tests and tell you whether the winner is real. I show every rate, the lift and how sure you can be, worked out from your own counts. Then I give you a plain call: ship it, keep running, or start again.

What Harriet delivers

A readout of a finished A/B test: Rates, lift, a 95% range and a p-value with the working shown, set-up checks and a plain ship, keep running or rerun call.
A check on a test that is still running: Whether it is safe to stop now, and how many more visitors or days you need, with the sums.
An honest read of a before-and-after change: What a change with no control group can and cannot tell you, with the numbers and a better test to run next.
A readout of a test with three to five versions: Each version against the original with a correction for testing many at once, and which one to keep.
A plan for the next test, with sample size: The hypothesis, what to change, sample size and length worked out, and stop rules set in advance.

Get Harriet, free

Download Harriet's pack from this page, or connect your AI and ask it to load Harriet. Read the personality in soul.md.