Customer interviews that find your wedge are different from the customer interviews most founders run, which mostly produce polite validation and no clarity. A wedge is the single sharp use case where your product is so obviously the right answer that the buyer needs no convincing. You do not invent a wedge in a planning doc. You hear it, usually as an offhand sentence in an interview where someone describes a problem with more heat than the rest of the conversation. The skill is running interviews that surface that sentence instead of burying it under feature feedback.

This post is the concrete version: how to structure interviews to find a wedge, the questions that work versus the ones that lie, the signals that mark a real wedge, and where to find these conversations already happening in public.

Key takeaways

• A wedge is the one use case where the buyer needs no convincing; you discover it, you do not design it.

• Most customer interviews fail because they ask about the future and opinions; wedges live in past behavior and emotion.

• The wedge signal is intensity - the problem they described with frustration, a workaround they built, or money they already spent.

• Interview both wins and losses; lost prospects often reveal the wedge more sharply than happy customers.

• People describe their wedge unprompted in public every day; those posts are unfiltered interview data.

Why do most customer interviews fail to find a wedge?

Because they ask the wrong kind of question. "Would you use a tool that does X?" and "what features do you want?" generate hypothetical, agreeable answers that point nowhere. A wedge is not a feature request. It is a specific, painful, recurring situation the person already has, and you only find it by digging into what they actually did, not what they think they would do.

The second failure is interviewing only happy customers. They confirm what already works. The sharpest wedge signals often come from people who almost bought and did not, or who churned, because their frustration is specific and unguarded. For the broader validation frame, see how to validate your SaaS idea with real buyers.

What questions actually surface a wedge?

Questions about the last specific time the problem happened, not about preferences or the future. The Mom Test principle applies: people lie to be nice when you ask about opinions, and tell the truth when you ask about their past behavior. Anchor every question to a concrete recent instance.

• "Walk me through the last time this problem actually happened." Specifics expose the real shape of the pain.

• "What did you do about it?" A built workaround or money already spent is the strongest possible wedge signal.

• "What did that cost you - time, money, or a deal?" Quantified pain marks a wedge worth building on.

• "Why hasn't this been solved already?" Reveals whether the gap is real or just unaddressed by accident.

• Then stay silent. The wedge often arrives in the unprompted sentence after you stop talking.

Avoid "would you", "do you think", and any question that lets them imagine a future. The wedge is in the past tense.

How do you recognize a real wedge when you hear it?

The strongest single signal is a workaround they built themselves. Effort spent on a hack is revealed preference; it proves the problem is acute and unsolved in their context. According to customer-development thinking popularized by sources like Y Combinator's startup library, the most reliable signal of a real problem is what people already do about it, not what they say they want. One sharp wedge heard from three people beats fifty feature requests. For converting that into outbound, see how to write an ICP for outbound.

Where can you find wedge conversations already happening?

In public, unprompted, all day. People describe their wedge on Reddit and LinkedIn without a founder in the room asking leading questions - venting about a workaround, asking if a tool exists, complaining about what they overpay for. That is interview data with the social-desirability bias removed, because they were not talking to you.

Reading those at scale by hand is impractical, which is where repco.ai fits. It is an AI sales rep that watches Reddit and LinkedIn for people describing the problem you solve, scores how strong the intent is, drafts a message tied to that specific post, and runs the follow-up from your own account. Every one of those posts is both a wedge signal and a live buyer, which is the rare case where research and pipeline are the same activity. See how to monitor Reddit for buying intent and the signal-based selling playbook for 2026.

Frequently asked questions

How many interviews do I need to find a wedge?

Fewer than founders expect if the questions are right. A sharp wedge usually shows up as the same intense story repeating across five to ten well-run interviews. Twenty interviews of "would you use this" produce noise; ten interviews about past behavior produce a wedge.

Should I interview customers or non-customers to find the wedge?

Both, but weight non-customers and lost deals heavily. Happy customers confirm what works; people who did not buy or who churned describe the gap with the unguarded specificity that exposes where your product is undeniably the answer.

What if every interview points to a different problem?

That usually means the questions were too broad or the segment was mixed. Tighten to one persona, anchor every question to the last specific occurrence, and look for the problem with the most emotional intensity. The wedge is the one with heat, not the longest list.

Bottom line

Customer interviews that find your wedge work when you ask about past behavior instead of future opinions, interview losses as hard as wins, and listen for intensity - a workaround built, money spent, a story repeated. The fastest source of that signal is people describing the wedge in public, unprompted, where research and a live buyer are the same conversation. Start at repco.ai.