Intuition
A detective doesn't dust every surface in the city for fingerprints. They form a theory — "the butler did it" — and then go looking for the one piece of evidence that would confirm or destroy it. Hypothesis-driven problem solving is the same move: instead of analyzing everything, you commit to a likely answer early and aim your limited time at testing it.
This is what separates a consultant from a data dump. Anyone can list ten things to look at. A consultant says which one probably matters and why.
Framework
- State a hypothesis early. "My leading hypothesis is that the profit drop is a cost problem, specifically rising logistics."
- Make it testable. A good hypothesis names the data that would prove it right or wrong.
- Test, then update. Go to the branch, get the number, and say what it means. Confirmed → go deeper. Killed → pivot, out loud.
- Stay disciplined, not stubborn. Hold your view confidently but drop it the instant evidence says otherwise.
Worked Example
A retailer's margins are shrinking. Weak approach: "Let me look at revenue, costs, competition, pricing, and the market." Strong approach: "Margins, not revenue, are the issue — so my hypothesis is that costs rose faster than price. Within costs, I'd bet on input prices given recent commodity moves. Can I see cost of goods over the last three years?" If COGS is flat, you say so and pivot to price erosion. You've spent your minutes like a detective, not a vacuum cleaner.
Pitfalls
- Refusing to commit — "I'd need to see all the data" reads as no point of view.
- Clinging to a dead hypothesis after the data killed it.
- A hypothesis so vague ("something is wrong with the business") that no data could test it.