Assumptions & Hypotheses

Write down what you believe but have not proven, say in advance what would prove each belief wrong, and put the rival explanations on the page before the data has a chance to flatter your favorite.

Mixed research: hypothesis-driven work comes from falsification (Popper, 1935) and consulting practice (Rasiel, 1999), with experimental support in entrepreneurship (Camuffo et al., 2020) but unproven transfer to internal investigations. The biases in Step 5 rest on strong research (Wason, 1960; Tversky & Kahneman, 1974). The confidence scale is a rule of thumb.

Use it when:

• The question is framed and people already have strong views about the answer.
• You are about to design interviews, a survey or an analysis and want it to be able to prove you wrong.
• A decision rests on a belief nobody has checked, such as why customers leave or why a process is slow.

Skip it when:

• You have no clear question yet: start with 2.1 Problem & Question Framing.
• You are exploring a field you know little about and cannot yet guess an answer: go straight to open interviews with 2.7 Data Capture Templates, and come back here when patterns appear.

A belief becomes a hypothesis only when you can say in advance what would show it to be wrong (Popper, 1935). Consultants turned that stance into a habit, organizing work around an early answer to be tested rather than a topic to explore (Rasiel, 1999), and product teams did the same with assumptions about users (Gothelf and Seiden, 2012). Whether this helps in business is mixed: a field experiment with startups supports teaching people to test their assumptions (Camuffo et al., 2020), but nobody has shown the same for internal investigations. The reasons the discipline is needed are better documented. People test ideas by looking for confirmation (Wason, 1960), lean on early numbers and on whatever comes easily to mind (Tversky and Kahneman, 1974), and draw lessons from the cases that survived to be seen (Mangel and Samaniego, 1984).

So the hardest field here is "What would disprove it", under every assumption. Step 3 puts rivals on the page before data arrives, Step 4 asks what an empty result would mean, and Step 5 turns the known biases into design choices instead of good intentions.

Write what the team believes but has not proven. For each one, give the evidence you have, what would prove it wrong, and how confident you are (1 low to 5 high, rule of thumb).

Strong answer: "Small cafes leave because our minimum order is too large. Evidence: most of their orders sat at the minimum. Disproved if: cafes that left usually ordered well above it. Confidence: 3." Not: "Customers want better value."

Illustrative example: the business, the roles and all numbers are made up.

The Gothenburg bakery from 2.1 is asking why small cafes stop ordering within six months.

Step 1: Three beliefs. Price is too high (the drivers' view; evidence: two cafes said so; confidence 2). The minimum order is too large (the sales lead's view; most orders sat at the minimum; confidence 3). Delivery comes too late for breakfast (one complaint; confidence 1).

Step 2: The minimum-order belief became the main hypothesis: cafes stop because they throw away bread. Confirmed if most former customers raise waste unprompted and current customers rarely do; rejected if both groups mention it equally.

Step 3: Rivals: owner or menu changes (likelihood 3), a new bakery nearby (likelihood 2).

Step 4: If former and current cafes look alike, the cause is outside the offer; next step is to check owner changes.

Step 5: Availability and selection were ticked. Guards: interview a random draw of former customers, and ask "why did you start ordering?" before any question about price or waste. The head baker, who thinks it is all about price, read the page and added the price rival to the interview guide.