Finalizing

A test isn't finished when the numbers come in; it is finished when someone decides what to do about them. Finalizing records that decision on the experiment — the outcome, what you will do and why — and freezes the results it was based on.

When you can finalize

Finalize appears in the page header once the experiment is Ready for decision: the test window has ended and the results cover all of it. If an evaluation is running, wait for it to finish, so the decision is based on the final results.

Recording the decision

Click Finalize and fill in the form:

Field	Options
Outcome	Win, Loss, No clear winner, Inconclusive
Winning variant	Only for a win: the variant that won
Decision	Roll out, Iterate, Re-test, Stop
Roll-out date	Only for Roll out: the day the change went live, today by default
Summary	What happened and why you decided this way

The form opens with the outcome, winning variant and decision Mitzu suggests from the results (see the table below). They are suggestions: the decision is yours, and it doesn't have to match the verdict. A significant win on a metric that turned out to be the wrong one can be recorded as Inconclusive.

The summary is drafted from the results, in sections you can edit:

• Outcome — the outcome you picked.
• Result — the verdict's headline and sentence.
• Rationale — the hypothesis and each variant's value on the primary metric.
• Segment observations — No segments were analysed.; results by segment aren't available yet.
• Behavioural findings — each secondary metric and guardrail, with every variant's lift and outcome.
• Limitations — every health check that warned or failed.
• Recommendation — a sentence for the decision, for example Roll out Short form to everyone.

Add the context only you have: why the test was run, what else was going on, what happens next. Markdown is supported. Redraft from results replaces your edits with a fresh draft, for example after you change the outcome.

Roll-out date appears when the decision is Roll out and starts on today. On finalize it adds a roll-out annotation on that day, named after the winner, so every chart shows when the change went live. Set it to the day the winner reached everyone in your tool, or clear it to skip the annotation.

Add the test window to annotations, ticked by default, marks the test as a date range on every chart in the workspace. See Follow-ups.

Click Finalize. Experiment finalized. The results are frozen.

What finalizing does

• The experiment becomes Finalized.
• The results shown are frozen: later evaluations never replace them.
• A Summary card shows the outcome, for example Win: Short form, the decision, for example Decision: Roll out Short form, and your summary, with a button to copy it.
• A Follow-up card appears, for saving variant cohorts and adding annotations.
• Duplicate as follow-up, Reopen and Re-evaluate appear in the page header.
• The experiment can no longer be edited.

Choosing the outcome and decision

What Mitzu suggests for each verdict, and when you might choose otherwise:

Results	Suggested outcome	Suggested decision	Consider instead
Significant win, guardrails within limits	Win	Roll out
Significant win, a guardrail breached	Win	Iterate	Roll out if the harm is small and understood
Significant loss	Loss	Stop	Iterate if the idea is still worth a better version
No clear winner	No clear winner	Stop	Roll out if the variant is cheaper or simpler to maintain
Inconclusive: sample too small, or too few users to compare	Inconclusive	Re-test	Stop if the lift you could detect isn't worth a longer test
Verdict withheld	Inconclusive	Re-test	Fix the assignment first

"No clear winner" is a result, not a failure: with a sample plan reached, it says the change doesn't move the metric by the lift you cared about. That is worth knowing and worth recording.

Writing a useful summary

A summary is read months later by someone who wasn't there. Cover:

1. What was tested and why — one sentence, from the hypothesis.
2. What happened — the verdict on the primary metric with the lift and interval, and anything notable on the secondary metrics and guardrails.
3. What you decided and why — including what the numbers alone don't say: cost, risk, strategy.
4. What happens next — the roll-out date, the follow-up test, the open question.