Miatz playground

The Peek

Stop an A/B test early, then watch it turn out wrong.

Coming to the Build-LabBrowser (JS)Datadata
The Demystify signature

running the same zero-effect test 100 times shows peeking-and-stopping calls a false winner far more often than the 5% everyone assumes — the real false-positive rate, counted live

How it works

What goes in, what comes out

What it does

The learner runs a simulated A/B test with a real, hidden effect size and can check the live p-value and stop the test at any point. A ghost trace then replays what would have happened had they run to the pre-registered sample size, showing how often peeking flips the declared winner.

You bring

watch the live results feed and decide when, or whether, to stop the test

You get

a p-value-over-time curve, the learner's called winner, and a ghost trace of the 'ran to completion' outcome for comparison

You control

true effect size (including zero — no real difference), traffic rate, pre-registered sample size, number of repeated simulations

Where it's used

Data L3 'Evals as the Data Team's Contract' crossover; Business and Marketing A/B modules

statistical significancethe peeking problemearly-stopping biassample size and power

Concepts click when you open the machinery.

Three labs are already live and free — the same hands-on style this playground brings to its module.