intermediate / August 2026

Experiment Decision Rules Intermediate Quiz

Experiment decisions need clear rules before a test is run, paused, or interpreted. Review hypotheses, sample limits, success metrics, random assignment, checks, side effects, and inconclusive results.

Before you start

Start a 10-question practice round.

Sign in before starting if you want a leaderboard score.

New

Sign in to get ranked
Questions
10
Time limit
7 min
Scoring
First signed-in attempt counts
Edition
August 2026

What this quiz checks

Decide what the test can prove before calling a winner

Experiment designSuccess criteriaSample planningRandom assignmentResult interpretation
  • Start With A Decision The Test Can AnswerA test should compare a clear change against a fair control. It is weaker when the team has not named the user behavior, success measure, or decision it will make.
  • Check Whether The Test Can Produce Useful EvidenceA fair design still needs enough traffic, time, and checks. Low volume or broken assignment can make the result too weak for a rollout decision.
  • Interpret Results With BoundariesAn experiment result is evidence for the tested change under the tested conditions. It should not be stretched into a wider claim than the design supports.

Study first

Review the ideas behind the questions

Check the experiment plan before traffic is split. Focus on the decision the test must answer, whether the comparison is fair, and what the team will do if the evidence is weak.

Start With A Decision The Test Can Answer

A test should compare a clear change against a fair control. It is weaker when the team has not named the user behavior, success measure, or decision it will make.

  • A/B testing compares two versions of a page or service to see which performs better against a defined measure.
  • A useful test-and-learn plan names the intervention, comparison group, outcome, and how the result will inform action.
  • The team should choose measures that show whether the service is solving the problem, not only whether activity increased.

In Practice

Name The Decision

Before launch, write the decision the result will support, such as whether to change a sign-up step, rewrite a price explanation, or continue research.

Pick One Main Outcome

A test can track several guardrails, but the main result should be tied to the problem the change is meant to solve.

Common mistakes

  • Running a test because a page feels weak, without naming the change or outcome.

    Define the intervention, comparison, outcome, and decision before splitting traffic.

Q&A

What makes an experiment different from a normal page change?

It compares a defined change with a control so the team can judge the effect more fairly.

Why decide the primary measure before launch?

It prevents the team from choosing the best-looking metric after seeing the result.

Check Whether The Test Can Produce Useful Evidence

A fair design still needs enough traffic, time, and checks. Low volume or broken assignment can make the result too weak for a rollout decision.

  • Teams should consider how many users visit each month and how many are needed before deciding whether an A/B test is appropriate.
  • Random assignment and a comparison group help make the comparison fair.
  • A test plan should check for practical delivery risks before the test starts.

In Practice

Low Traffic Changes The Method

If only a small number of people reach the step each month, a usability study, customer research, or staged rollout may answer the decision better than a formal A/B test.

Evidence Needs A Checking Plan

A variant that fails on one device or sends the wrong users into the wrong branch can create a measurement problem, not just a delivery bug.

Common mistakes

  • Ignoring sample limits because the test tool can technically launch.

    Check whether the expected traffic and sample can support the decision before launching.

Q&A

What should a low-traffic team check before launching a test?

Check whether enough users will enter the experiment to support the decision in the available time.

Why does assignment quality matter?

The comparison is weaker if the groups are not comparable or the branch logic sends the wrong people into a version.

Interpret Results With Boundaries

An experiment result is evidence for the tested change under the tested conditions. It should not be stretched into a wider claim than the design supports.

  • Test results should be interpreted against the planned outcome and decision, not only the highest-moving metric.
  • A test-and-learn review should consider unintended consequences, not only the main outcome.
  • Performance data can show a problem or effect, but teams may still need user research to understand why it happened.

In Practice

Do Not Move The Goal Afterward

If the main measure did not move, a secondary lift can be useful learning, but it is not the same as winning the planned test.

Look For Side Effects

A signup lift can still be a bad rollout if support contacts, refund requests, task failures, or accessibility issues rise at the same time.

Common mistakes

  • Calling a test a winner because one unplanned metric improved.

    Treat the unplanned lift as learning and check it against the planned primary outcome and guardrails.

Q&A

What should happen when the result is inconclusive?

Keep the decision modest, review whether the test had enough evidence, and choose research or another test if the question still matters.

Why check user research after a test result?

The metric may show what changed, while research can help explain why users reacted that way.

Question quality

Reviewed before publishing

Reviewed by
Aniruddh Sharma
Last checked
August 22, 2026

Reviewed against GOV.UK A/B testing, test-and-learn, and service performance guidance because this quiz tests platform-neutral experiment readiness, sample limits, fair comparison, and result interpretation.

The source pages for this edition were checked as part of the same review. Official product docs are linked where available.

Sources

Sources used for this quiz

These pages support the quiz content and study notes.