Test Hypothesis
A test hypothesis is a clear, testable statement predicting that a specific change will improve a defined metric for a defined reason.
In depth
A strong hypothesis follows a structure such as "Because we observed X, changing Y will cause Z, measured by metric M." Anchoring the prediction to an observation forces you to base experiments on evidence rather than opinion, and naming the metric in advance prevents the cherry-picking that quietly inflates win rates. The mechanism clause also matters: it captures the reasoning you can validate or reject, so even a losing test produces durable learning.
In a quiz-funnel context, hypotheses keep optimization disciplined when many small levers compete for attention. A team might hypothesize that reducing a scorecard from twelve questions to seven will lift completion because friction, not interest, is causing drop-off. The common pitfall is writing a hypothesis so vague it can never be falsified, which makes the result un-actionable; tie every experiment to one primary metric and a plausible cause so the funnel improves through understanding rather than guesswork.
Example in practice
Frequently asked questions
What makes a test hypothesis good?
A good hypothesis ties a change to an observed problem and predicts a single primary metric. It states the expected mechanism, so even a losing test teaches you something about your audience.
How is a hypothesis different from an idea?
An idea is just a proposed change, while a hypothesis predicts a measurable outcome and a reason. That predictive, falsifiable structure is what lets you learn from the result rather than simply ship the change.
Do I need a hypothesis for every test?
Yes, every experiment should start with one because it forces you to define success before you see the data. Without it, teams unconsciously redefine winning to match whatever the test produced.