An experiment hypothesis from interviews: problem, change and metric
A useful hypothesis links a specific problem, a proposed change and an observable outcome. It says up front what the team will learn and what outcome will lead it to keep, change or stop the intervention.
How to write a testable hypothesis after interviews
A useful hypothesis links a specific problem, a proposed change and an observable outcome. It says up front what the team will learn and what outcome will lead it to keep, change or stop the intervention.
A transcript is a verifiable source of wording. It preserves context, qualifiers and the order of the reasoning. The author still needs to decide on the final wording and test it with participants.
Build the hypothesis step by step
Record the evidence
Note the problem, segment and timecode. A single striking quote gives you a lead, not a general conclusion.
Choose a change
Describe the smallest change to the experience that you can test. Keep the scope focused.
Set a metric
Specify the user action, observation period and baseline if you have one.
Define the decision
Before the test, write down what you will do if the result is strong, weak or inconclusive.
Example: a reminder about a transcript
This is a fictional training example. The names, circumstances and outcomes are included solely to explain the method.
“I upload a file, get distracted and remember to check the result the next day.”
The evidence points to a gap between processing finishing and the user coming back. One change to try is a single notification when the transcript is ready. A redesign of the account area is unnecessary for this test.
Hypothesis: if users who left the page receive a clear notification, a larger share will open a finished result within 24 hours. Set the threshold and duration using the team’s own data. The interview does not provide a baseline.
| Field | Question | Example |
|---|---|---|
| Segment | Who experiences the problem? | User with a long recording |
| Change | What changes? | One notification when it is ready |
| Metric | What will we observe? | Opening the result within 24 hours |
| Decision | What will we do afterward? | Keep, change or stop |
Experiment hypothesis template
Copy the template and fill it in using a specific recording. Label guesses as hypotheses.
Problem we observed: Segment affected: Change to test: Expected action: Observation period: Decision criterion and limitations:
Four checks before sending
- Evidence and interpretation are recorded separately.
- You can test the change before building the full solution.
- The metric reflects behavior.
- The response to each possible outcome was decided before the test.
A closer look
Do we need a numeric threshold?
Yes, if the decision depends on a number. Choose the threshold using the baseline and acceptable risk.
Can we test several changes?
Yes. It will be harder to tell which one caused the result. One main change is easier to interpret in the first round.