🗣️ Customer Discovery

Why do it — and why almost everyone does it wrong

Customer discovery is the step that separates the founders who build what the market wants from those who build what they imagine the market wants. In practice, almost everyone does it — and almost everyone does it wrong.

The classic mistake: you call ten people you know, explain your idea, ask "would you use it?", collect ten enthusiastic answers, and declare the hypothesis validated. You haven't discovered anything. You've confirmed that people are polite and that your pitch is compelling enough to extract a "yes" from friends.

Real customer discovery works the opposite way. You don't explain your idea. You explore the customer's problem. You listen to behaviour, not opinions. You ask about the past, not the future. And you exit the interview with data you can act on — not encouragement.

What is customer discovery

Customer discovery is the structured process of understanding your potential customer's problem deeply enough to design a solution they would actually use, pay for, and recommend. It's not a one-time activity — it's a discipline that runs through the entire early phase of a startup.

The term was formalised by Steve Blank in "The Four Steps to the Epiphany" and operationalised through the Customer Development methodology. The core insight: before you build, you need to become the world's foremost expert on your customer's problem. Not on solutions — on the problem.

How to use it: the interview protocol

Step 1: Define the hypothesis you're testing
Every interview has a precise objective. Before each conversation, write down: what specific assumption am I trying to validate or invalidate today? Don't enter an interview to "learn generally" — you'll exit with scattered impressions. Enter with a hypothesis and exit with a verdict.

Step 2: Find the right interviewees
The right people to interview are those who have the problem you're solving — not those who would hypothetically benefit from your solution. The difference matters. If you're building a B2B tool, talk to the person who actually does the job, not the VP who signed off on buying tools. If you're building a consumer product, talk to people with the specific behaviour pattern you're targeting.

Step 3: Ask about behaviour, not opinions
The most important principle of the Mom Test (Rob Fitzpatrick): never ask "would you use X?" Ask "tell me about the last time you tried to solve Y." Opinions about hypothetical futures are unreliable. Descriptions of past behaviour are data. The interview should be 80% the customer talking, 20% you asking follow-up questions.

Step 4: Listen for signals, not confirmations
Strong discovery signals include: the customer describes the problem unprompted in the exact terms you'd use; they show you the workarounds they've built; they ask "can I try this?"; they name three colleagues who have the same problem. Weak signals: "interesting idea," "I could see using this," "someone might pay for that." Learn to distinguish the two.

Step 5: Extract operational insights
After each interview, write down: what was the most surprising thing? What changed in my understanding of the problem? What assumption did this interview invalidate? What do I need to test next? Discovery is a learning loop — each interview should update your model, not just confirm it.

3 common mistakes

1. Pitching instead of listening
The moment you explain your solution, you've broken the interview. People will now respond to your idea, not describe their problem. Keep the solution out of the conversation until you've exhausted the problem space.

2. Interviewing the wrong people
Interviewing people who are "probably" in your target market produces weak signal. The person who has actively tried to solve the problem, built a workaround, or complained loudly about the status quo — that's who you need to find. Five interviews with the right person beat fifty with the wrong one.

3. Not taking notes on surprises
Most founders exit interviews remembering what confirmed their thesis. The most valuable information is what contradicted it. Keep a "surprise log" after each interview. Those surprises are where the real insights live.

Real Examples

Intercom (Ireland): Before building their messaging platform, the Intercom team spent weeks interviewing SaaS companies about how they communicated with users. The key discovery: the actual pain wasn't "we need a chat widget" — it was "we have no idea who our users are or what they're doing." That insight shaped Intercom's focus on user identity and behaviour tracking, which became the company's core differentiator.

Pipedrive (Estonia): The Pipedrive founders, themselves former salespeople, interviewed dozens of sales reps before writing a line of code. Their discovery: existing CRMs were built for managers who wanted reporting — not for salespeople who needed to manage their pipeline visually. That user-side insight, gleaned from customer discovery, created a product category: the sales-rep-first CRM.

Algolia (France): Before building a search API, the Algolia team interviewed developers about their experience implementing search. The consistent finding: developers hated building search from scratch but also hated Elasticsearch's complexity. The specific pain — "I want great search without becoming a search expert" — defined both the product and the go-to-market.

Recommended Resources

The Mom Test — Rob Fitzpatrick — the definitive guide to customer interviews that produce real signal. Read this before your first interview. — https://www.momtestbook.com/

Talking to Humans — Giff Constable — practical guide to qualitative research for startups, freely available online. — https://www.talkingtohumans.com/

The Four Steps to the Epiphany — Steve Blank — the foundational text for Customer Development methodology. — https://www.amazon.com/Four-Steps-Epiphany-Steve-Blank/dp/0977999300

Next Step with IdeaLedger

Once you've run at least 15-20 interviews and identified 3 repeating patterns, you're ready to move to Step 5 (Buyer Persona). Don't build the persona from guesses — build it from the patterns you found in discovery. If you haven't found consistent patterns yet, keep interviewing.