Start with an observation
Write down what you noticed before trying to explain it. Separate the observation from the story you tell about it. “Visitors leave this page quickly” is an observation; “the design is bad” is one possible explanation.
Make the question testable
Specify the population, the outcome, and the comparison. Instead of asking whether a tool is better, ask which task it improves, for whom, and under which conditions. Name what you can measure and what you cannot.
Choose evidence before the answer
Decide what result would support your explanation and what would weaken it. This makes it harder to quietly move the goalposts after you see the data. Keep a record of alternative explanations.
Leave room to revise
A small, well-defined question can be more useful than a sweeping claim. Document uncertainty and say what the next experiment should distinguish. Understanding grows when a question becomes more precise.