Design an experiment readout your team can reuse
Record enough context to avoid repeating inconclusive tests.
The scenario
A checkout experiment showed a 6% lift, but nobody recorded the observation period or who was included.
A useful readout documents sample, allocation, metrics, uncertainty, and any implementation errors, making the result easier to interpret later.
Apply it to your own work
Save the hypothesis, version details, dates, counts, decision, and next question. Add screenshots or links where they explain the difference.
What to watch for
A result is not automatically portable to another audience or season. Describe where the finding applies.
Make uncertainty visible in the readout
Include counts by arm, the definition of an eligible outcome, and an uncertainty interval where the design supports one. Record dates and any outages, tracking changes, or unusual traffic splits.
Write down what the team will do if the evidence is inconclusive. A follow-up can target a smaller decision or fix instrumentation. Keep the original result accessible so the team does not interpret the same data differently after the fact.
Try a related tool
Find the share of eligible visits that completed a defined action. Count at most one conversion per visit for this calculation.
