A hypothesis is a clear, testable proposition about expected outcomes, relationships, or predictors. In analytics, it turns an operational question into a basis for choosing research, data-gathering, and evaluation methods. section 3.0 section 9.2
Construct a strong hypothesis by making its decision criteria explicit:
- Identify the independent variables or predictors and the dependent variable or outcome. section 3.1.2
- State the expected outcome declaratively and clearly enough for stakeholders to interpret. section 3.1.2
- Specify a directional expectation when appropriate, and quantify the expected difference when useful for evaluation. section 3.1.2
- Document the evidence, missing information, constraints, and reasoning behind the proposition rather than waiting for perfect context. section 3.1
Tutoring and test scores
H0: Individual tutoring produces test scores equal to or lower than no tutoring. H1: Individual tutoring produces higher test scores than no tutoring. Results support rejecting H0 only when they meet the predefined directional and statistical criteria. section 3.1.2
E-commerce experiment
An analytics team can test three page layouts against a control layout to reduce abandoned carts, then state null and alternative hypotheses for the group differences and quantify the expected outcome. section 3.1.4
Figure 3.1 When you and your stakeholders agree on the question, hypothesis, and methods, you’ll be able to better focus on delivering concrete recommendations.
Use the figure to connect agreement on the question, hypothesis, and methods with focused recommendations for stakeholders. section 3.1
