Customers & Research

I've done 20 interviews and every person described a slightly different problem. How do I know if there's a real pattern or just noise?

A starting point

Scattered problems usually mean your target group is too broad, not that there's no signal. As a starting point, sort the transcripts by who the person is and how badly they hurt, and look for the small cluster that describes the same pain in the same words. One narrow, intense, repeated problem beats twenty vague ones, so pick the sharpest cluster and go deeper before you widen again.

Go deeper

Hand-picked from around the web, each with a note on why it earns your time.

3 resources 3 link-checked Read Use

Read

📖 Book
✓ Link checked Freemium Intermediate

Why we picked it When 20 interviews each surface a slightly different problem, the fix is not more interviews, it is a structure to sort them. Teresa Torres's opportunity solution tree gives you exactly that: a way to map every pain point you heard against one desired outcome, so real, recurring opportunities separate themselves from one-off noise. This guide is the clearest free explanation of the method, and it points to the full book if you want to go deeper.

Continuous Discovery Habits: Discover Products that Create Customer Value and Business Value

From Product Talk by Teresa Torres ~10 min read (companion to a full book)

  • Group each interview finding as an opportunity (a customer need, pain, or desire) and place it under a single desired outcome, so patterns become visible instead of a flat list of quotes.
  • A problem that keeps recurring across interviews earns a spot on the tree, a one-off mention does not, which is a concrete rule for pattern versus noise.
  • The tree is meant to keep changing as you keep talking to customers, so treat your 20 interviews as a starting point, not a finished map.
Open producttalk.org →
📄 Article
✓ Link checked Free Intermediate

Why we picked it Turns the fuzzy phrase 'product-market fit' into a number you can move. Vohra's survey (the '40% would be very disappointed' benchmark) and 4-step engine took Superhuman from 22% to 58% PMF. The most actionable PMF piece on the internet.

How Superhuman Built an Engine to Find Product/Market Fit

From First Round Review by Rahul Vohra (CEO, Superhuman) ~25 min read

  • Ask users: 'How would you feel if you could no longer use this?' Target >40% 'very disappointed'.
  • Segment to your highest-expectation users and build for them.
  • Split your roadmap: double down on what they love, remove what blocks the fence-sitters.
Open review.firstround.com →

Use

🛠️ Tool
✓ Link checked Freemium Beginner

Why we picked it Reading 20 transcripts in your head is where patterns get lost. Dovetail lets you import interview transcripts, tag the moments that matter, and cluster those tags into themes through affinity mapping, so a problem that shows up in eight interviews stops looking like eight unrelated complaints. It is a practical way to turn a pile of conversations into a countable set of recurring issues.

Dovetail

From Dovetail by Dovetail Ongoing tool, free tier available

  • Tag interview transcripts by the underlying problem, then let the affinity map group those tags so recurring themes surface on their own.
  • Seeing how many interviews carry the same tag turns 'this feels common' into an actual count, which is the difference between a pattern and noise.
  • You can start on the free tier with your existing transcripts, so it is worth trying before you commit to paid research software.
Open dovetail.com →

People also ask

How do I talk to customers without them just telling me what I want to hear? Stop pitching and start interrogating their past. Ask about what they actually did last time the problem hit, what it cost them, and what they trie... Beginner 3 resources → What questions should I actually ask in a customer interview? Ask about the past and the concrete, not the future and the hypothetical. 'Walk me through the last time you dealt with X', 'what's the hardest par... Beginner 3 resources → How many customer interviews do I need before I trust the signal? There's no magic number, you're done when you can predict roughly the next answer before they say it. In practice that's usually 15-30 conversation... Intermediate 2 resources → Where do I find people to interview when I have zero audience? Start with the person you already know who has the problem, then ask each interviewee for two intros. Go where the customer already hangs out, nich... Beginner 2 resources → How do I run customer discovery as an Indian founder building for Indian users? The frameworks are global but the field is local: Indian buyers often say 'yes, nice' to your face and never pay, so watch what they do, not what t... Intermediate 2 resources → How do I show a prototype to users without biasing their reaction? Hand it over, give them a goal, and shut up. Don't narrate or explain, watch where they get stuck, what they ignore, and what they say out loud. Th... Intermediate 2 resources →