Skip to content
Skills from A Smart Bear

Hypotheses: Your best guesses, written down to facilitate learning

Convert your untested beliefs about customers, pain, and price into numbered, falsifiable hypotheses — the raw material your customer interviews are designed to test. Adapted from Chapter 10 of Hidden Multipliers: A multiplier of one's own.

Input
Your GOALS.md goal questions, plus what you currently believe about customers, pain, price, and competition.
Output
A HYPOTHESES.md file: numbered, falsifiable hypotheses (H1, H2, …), each mapped to the goal questions it addresses.

What this is about

Writing down answers before you interview anyone can feel wrong. The point of an interview is to find the truth, not to assume you already know it. But recorded predictions are what let interviews teach you something. An unwritten belief cannot lose an argument with reality. Confirmation bias will quietly ignore every answer that contradicts what you already think. Write down your best guesses first, even the ones that seem too obvious to write. This step defeats confirmation bias. It is also where good interview questions come from. Each question is a small experiment. Each experiment tests one hypothesis.

This is the second step of the interview workshop, which starts with goal questions. This skill takes your GOALS.md as input. It walks through your goals one at a time. For each goal, it draws out what you actually believe. Then it presses each belief into a specific, falsifiable claim. For example, "customers care about speed" is only a feeling. "Customers with over X page views lose revenue when the site is slow, and have paid to fix it" is a claim that can be wrong. When you get stuck, the skill offers candidate hypotheses. Use these only as templates to correct. The list must state your bets. Half the value of this exercise comes from the work of forming your own beliefs. A plausible list you did not own teaches you nothing when reality contradicts it. When the list is sharp, the skill saves it in HYPOTHESES.md. Each hypothesis is numbered H1, H2, and so on, and mapped to the goals it addresses. The list is then ready for the next step of the method.

Expect half of your hypotheses to be wrong. This does not mean your list is bad. It is the reason you interview customers.

Example invocation

You can invoke the skill like this:

/asb-interview-hypotheses Here's my GOALS.md from the goal-questions
exercise. Let's write the hypotheses — I have a lot of opinions about our
property-manager customers but I've never written any of them down.

The session walks through your goals one at a time. It draws out what you believe and turns vague hunches into specific, falsifiable claims. You end with HYPOTHESES.md. It states your unvalidated beliefs in prose, then lists them as H1, H2, and so on, mapped to your goals.

One practical note: state where you are working. Name a directory, or point to your existing files, when you invoke the skill. The skill then keeps the method's files together in that location. If you do not name a location, the skill asks before it creates anything.

From the source

Two sources form the foundation of this skill:

  • As in Chapter 10 of Hidden Multipliers: A multiplier of one's own (section "Hypotheses: Your Current Best Guesses") — hiddenmultipliers.com — the hypotheses step itself. It gives the prediction-science case for writing down guesses, the eighteen-hypothesis WP Engine example with its goal mappings, and the famous wrong guesses (security, free trials) that interviews later disproved.
  • The Iterative-Hypothesis customer development method — the published article version of the method. It warns that AI-generated hypotheses are only templates. You must correct them to match what you actually believe, because half the value comes from thinking it through yourself.

Supporting articles each supply a piece of the mechanism: