Put the plan beside the result
An experiment readout should begin with what the team decided before seeing the outcome. Give Nema the hypothesis, variants, eligible audience, assignment method, planned dates, primary metric, and stopping rule. Then provide the result export and the metric definitions used to produce it.
If the plan changed during the test, include that history. A later definition should not silently replace the one used to make the original decision.
Check whether the comparison holds
Ask Nema to confirm that both variants cover the intended period and population. It can flag uneven exposure, missing rows, assignment changes, and late events. It should show raw counts and denominators beside rates so the reviewer can see the scale of each result.
Suppose one landing-page variant reports a higher conversion rate but received much less traffic after a tracking interruption. The readout should describe the observed result and the interruption. It should not declare a winner from the percentage alone.
Keep statistics in scope
Statistical interpretation depends on the design. Give Nema the method your analyst approved rather than asking it to choose a test after seeing the numbers. The readout can reproduce the agreed calculation and surface missing assumptions. An analyst checks assignment, sample size, independence, and whether the test ran as planned.
Secondary metrics belong in their own section. They can explain tradeoffs, but they should not be presented as the original success measure when the primary result disappoints.
Prepare the decision
The final section can list supported observations, caveats, and next questions. The marketing owner decides whether to adopt a variant, keep testing, or stop. Record that choice and its reason in the Slack thread so a later review does not confuse the output with the decision.
If the result needs more mature conversion data, set a new review date. Nema can prepare the updated readout when the approved inputs arrive.
Start with a completed test
Use a past experiment with a known plan and clean export. Ask Nema to recreate the readout, then compare it with the analyst's original work. Differences will expose unclear definitions and weak input assumptions before the workflow touches a current decision.
Questions from growth teams
Does Nema decide statistical significance?
Nema can apply a supplied analysis method and show its inputs. An analyst should validate that the method fits the design and interpret the result.
What if the test stopped early?
State when and why it stopped. The readout should compare that event with the planned stopping rule and treat the result with the appropriate caution.
Can the review recommend a winner?
It can summarize evidence for the defined decision. The marketing owner chooses the action after reviewing method, impact, and operational context.