Customer discovery: process, questions and decision tree

Most early-stage risk falls into four groups, and discovery touches each one:

Interviews mostly address problem and customer risk. Prototypes and concierge tests address solution risk. Pricing conversations and pilot offers start on go-to-market risk.

Before the first interview, write down what must be true for the idea to work. Then rank each assumption by how important and how uncertain it is, and test the top one first. An illustrative map for a member retention tool:

Each line becomes a learning goal. Without the map, every conversation feels useful and none of them changes a decision.

Each round narrows the question. Six steps cover most of it:

1. 1

   Set a learning goal. 'Learn whether agency owners struggle to forecast project profit' is better than 'talk to agencies'.

2. 2

   Choose one segment. Narrow enough to see patterns: role, industry, company size, workflow or the moment the need appears.

3. 3

   Recruit people who have the problem. Professional groups, communities, referrals and outreach. Friends are good for practice, not evidence. Say you are researching a problem, not selling.

4. 4

   Interview about real situations. Ask about the last time the problem happened, what they did and what it cost.

5. 5

   Look for patterns. Group notes by trigger, workaround, buyer, language and objections after each batch of interviews.

6. 6

   Test the concept lightly. A landing page, clickable prototype, concierge service or pilot offer, and watch what people do.

Rob Fitzpatrick's book The Mom Test makes the core point: ask about people's lives and past behavior, not your idea. Some swaps:

Talk less than they do, and write down their exact words. The phrases customers use to describe the pain become your positioning later.

Not every problem you hear deserves a product. Rate each one on how painful it is, how often it happens, what people do today and how ready they are to change. An illustrative scoring for a reporting tool:

A requested feature is not the same as a need. A request for 'more customization' might really be about control, trust or fewer mistakes. Find the underlying need before you choose the product response.

A persona helps when it comes from interviews, not imagination. 'Marketing Manager Mary, 34, likes productivity tools' says little. Fill in these fields from what you heard:

Split a persona when the evidence says to, for example when the person who pays is not the person who uses the product.

Interest is not evidence. The stronger the commitment, the more weight the signal carries:

Expect many ideas to fail even after good discovery. In Harvard Business Review, Ron Kohavi and Stefan Thomke report that at Microsoft about one-third of tested ideas prove effective, with the rest neutral or negative. Behavior is the test that counts.

After each round, walk through these questions in order. A 'no' tells you what to work on next:

1. 1

   Does the segment clearly have the problem? If not, try another segment or revisit the problem assumption.

2. 2

   Is it painful, frequent or costly enough to act on? If not, look for a more urgent problem or a group with more at stake.

3. 3

   Are they already trying to solve it? If not, the problem may not be urgent yet. If so, study the workaround closely.

4. 4

   Do you know the buyer and how they decide? If not, interview the other people involved.

5. 5

   Does your concept beat the current alternatives? If not, refine the value proposition or test another approach.

6. 6

   Will customers take a meaningful next step? If not, look at trust, timing, price or urgency. If so, move to deeper tests, pilots or early sales.

Feature requests pile up fast. Before one goes on the roadmap, ask:

• From the target segment? Or from an edge case?
• Linked to a high-priority need? One you found in discovery.
• Changes a key behavior? Activation, retention, conversion or trust.
• The problem or one suggested fix? 'More templates' may mean people don't know how to start, which onboarding could solve.
• Heard from several customers? In different words, about the same need.

Small usability tests catch a lot of this early. Jakob Nielsen of the Nielsen Norman Group found that a first usability test with five users finds about 85% of the usability problems.

Most weak discovery comes from looking for confirmation instead of learning:

1. 1

   Pitching too early. Once you pitch, people evaluate your idea instead of describing their world.

2. 2

   Hypothetical questions. 'would you buy this?' produces answers people won't act on.

3. 3

   The wrong people. Opinions from outside the segment add noise.

4. 4

   Ignoring negative signals. No urgency, an unclear buyer or repeated confusion is evidence too.

5. 5

   Overbuilding before validation. A big build makes you defend it instead of learning from it.

6. 6

   No surprises. If every interview 'validates' the idea, the questions are probably leading.