A hypothesis is a specific, testable prediction about what change will improve a defined outcome. In B2B marketing and sales testing, a hypothesis takes the form: 'if we change X, then Y will improve because Z.' The change is the variable, Y is the metric you are optimising, and Z is the reasoning based on evidence or logical inference about why the change should matter.

Writing hypotheses forces discipline in testing. Without one, teams run tests without a clear basis for expecting improvement and interpret results without a framework for learning. With a hypothesis, every test produces a learning regardless of whether it confirms or refutes the prediction: either you have evidence supporting the reasoning, or evidence against it, and both are useful.

The quality of a hypothesis determines the value of the test. Weak hypotheses like 'changing the subject line will improve opens' generate weak learnings. Strong hypotheses like 'a subject line referencing the prospect's specific industry challenge will outperform a generic curiosity line because our audience responds to evidence of relevance over curiosity gaps' generate learnings that inform future decisions beyond the immediate test.

Analytics terms are useful only when they change a decision. A metric can look sophisticated and still be low value if nobody knows how it is calculated, which segment matters, or what action should follow when it moves. It usually becomes more useful when it is defined alongside A/B test, Iteration, and Baseline.