Hypothesis-Driven Analysis: Ask First, Analyze Second

WHY IT EXISTS: To give data analysis focus and avoid wasting time on irrelevant explorations or finding meaningless correlations. It makes decision-making more scientific by forcing you to test specific, falsifiable statements before acting.

THE MENTAL MODEL: Think of it as the scientific method applied to business data. You start with a clear, testable question ('Will this new button increase clicks by 10%?') and then collect data to get a definitive answer, rather than just 'looking at the data' for something interesting.

HOW IT WORKS: The process is structured. First, you state a clear, measurable hypothesis. Second, you identify the exact data needed to test it. Third, you collect and prepare the data. Fourth, you perform a statistical analysis to evaluate the evidence. Finally, you conclude whether the data supports or rejects your initial hypothesis.

WHEN TO USE IT: It's best when you have a specific question to answer or a decision to make. Use it for A/B testing product changes, performing root cause analysis on a metric that has changed, or validating a business strategy before committing significant resources.

WHEN NOT TO USE IT: Avoid it for initial, open-ended discovery where the goal is to generate new ideas or understand a domain broadly. In these situations, exploratory data analysis is more appropriate for finding interesting patterns that can become hypotheses later.

ONE CANONICAL EXAMPLE: A marketing team hypothesizes that emails with emojis in the subject line have a higher open rate. They create two versions of an email—one with an emoji, one without—and send them to two random segments of their user base. They then analyze the open rates for both groups to determine if their hypothesis was correct.