Analytics the Right Way

Chapter 6Making Decisions Through Hypothesis Validation

Performance measurement, which we discussed in the previous chapter, is inherently backward-looking: it’s about the expectations for an investment that were set in the past, and then comparing those expectations to actual business results in the present. Hypothesis validation, on the other hand, is inherently forward-looking. It’s about making the best decisions possible given imperfect, incomplete data and a finite amount of time.¹

A useful starting point for exploring hypothesis validation is with a formal definition of the word “hypothesis”:

Hypothesis: A tentative assumption made, in order to draw out and test its logical or empirical consequences.

Consider the three bolded elements in this definition. The first is the tentative assumption. A hypothesis starts with an idea. It does not start with data. While it’s possible that, in the business context, a particular chart or report might spark an idea, it’s also possible that raw critical thought—an instinct, or an observation of a competitor’s behavior, or a passing comment by a colleague, or a conversation with Richard Feynman in the middle of a lucid dream, or something else—might spark an idea. Isaac Newton observed an apple falling out of a tree, wondered why it was that the apple fell to the ground rather than falling upward or sidewise, and hypothesized that ...