Customer Discovery vs Customer Validation: What Changes Before and After PMF

•7 min read

A lot of teams run ten demos, hear "looks great" ten times, and tell their investors the idea is validated. Nothing was tested. Customer discovery finds out which problem is worth solving and for whom. Customer validation tests whether a specific solution, at a specific price, gets bought and used. They use different questions, produce different evidence, and fail in different ways. Product-market fit changes the mix between them; neither one ends.

Two jobs, in a fixed order

The split comes from Steve Blank's customer development model in The Four Steps to the Epiphany. Discovery and validation are the first two steps, in that order, and a failed validation sends you back to discovery rather than forward to scaling.

Discovery surveys the territory. Which problems exist, which segment has each one, what it costs them, what they do about it today. You end with a map: this problem lives in this segment and hurts this much, that one is loud but nobody owns a budget for it.

Validation picks one route across that map and tests it. This product, this segment, this price. You end with a yes or a no, and the evidence for it.

Skip the first step and you're testing a route across territory nobody surveyed. When the test fails, you won't know whether the solution was wrong or the problem was.

The questions change

Discovery questions are about their life, in the past tense: when the problem last happened, what it cost, what they pay today. You never pitch in a discovery call, because everything said after a pitch is a reaction to you instead of a report about them. The specific questions worth asking are in Customer Discovery Questions That Reveal Real Buying Intent.

Validation questions are about your offer, and they're concrete. Here's the product, here's the price, can we start with your team next month. In validation, the pitch is the experiment. Vague questions are the failure mode now, because a validation call that ends without a decision tested nothing.

The evidence changes

Discovery evidence is what people said about their own behavior: the missed deadline, the spreadsheet someone maintains, the invoice they already pay. It's words, so it needs receipts: the verbatim quote, the speaker, the timestamp, and a count of how many distinct people said it.

Validation evidence is what people did about your product: a paid pilot, a signed order form, usage that's still there in week four, an intro to the person who owns the budget. Words matter less here. A prospect who says "we'd definitely use this" and won't book the follow-up meeting has given you discovery data, and it reads "no."

Compliments count as evidence in neither mode.

The failure modes change

Discovery fails quietly, through leading questions. Ask people to agree that reporting is painful and they'll agree, and you'll walk away holding a market you manufactured on the call.

Validation fails through politeness accounting: logging "they loved the demo" as a yes. And the two failures compound. A team that skipped discovery and validates a solution to an unconfirmed problem gets the worst version: demos that sound great, deals that never close, and no evidence underneath to explain why.

You can audit which mode a past call actually was. Read the transcript and check two things: who talked more, and whether the questions were about their life or your product. A "discovery call" where you talked for thirty of forty-five minutes was a pitch meeting.

Before PMF: discovery leads, validation stays small and cheap

Pre-PMF, most of your conversations should be open discovery interviews, run in rounds per segment. How many you need is covered in its own article; the short version is 10 to 15 per segment, stopping when calls stop surprising you.

Validation bets stay small: a prototype in front of five people, a price said out loud at the end of a call, one concrete ask. What doesn't count is a survey asking "would you buy this," which is future-tense politeness collected at scale.

After PMF: validation leads, discovery changes aim

Post-PMF the mix flips. Most conversations now validate something specific: a feature, a price change, a new tier, a second segment.

Discovery re-aims. New segments you're expanding into get the same open interviews your first segment did. Churned customers get discovery calls, not save attempts, because "why did you actually leave" is a discovery question. You found fit with one market at one point in time, and markets move. Teams that stop discovering usually find that out from the churn numbers.

Where UpSight fits

Both modes run on conversations, and the difference between them shows up in the evidence. UpSight keeps every call: record with the desktop recorder, run built-in surveys, import audio, video and text files (mp3, mp4, vtt, srt, txt and more) from any tool, or have Otter send transcripts over automatically. Themes cluster across conversations and link to the verbatim quotes with speaker and timestamp, so "this problem is real in this segment" and "three pilots said yes" are both claims you can click into. Every answer points back to the quotes behind it, so you can check it.

Pricing is on the pricing page: Free at $0, Pro at $29 per month, Team at $39 per user per month.

FAQ

Is customer validation part of customer discovery?

No. In Blank's customer development model they're separate steps in a fixed order: discovery first, validation second. A failed validation loops back to discovery. Treating them as one activity is how teams end up pitching in interviews and wondering why the "validated" demand never converts.

Do I still need customer discovery after product-market fit?

Yes, aimed differently. New segments, churned customers, and expansion opportunities all need the same open, past-tense interviewing your first segment got. PMF describes your fit with a market at a point in time, and the market keeps moving after you stop looking.

How many discovery interviews before I start validating?

Stop discovering when two or three interviews in a row add nothing new to the themes you plan to act on, which usually lands in the 10-to-15-per-segment range. Validating earlier than that means testing a solution against a problem map you haven't finished drawing. The full stopping rule is in How Many Customer Interviews Is Enough?

Can a survey validate demand?

A "would you buy this" survey can't, because it collects future-tense answers that cost nothing to give. Surveys work for sizing a theme you already understand and for collecting facts about past behavior. Validation needs a real offer and a commitment that costs the buyer something.

What counts as validation evidence?

Behavior with a cost attached: money paid, a signed commitment, usage that persists past the novelty window, or someone spending their own credibility, like introducing you to their boss. If it cost the prospect nothing, it validated nothing.

Next step

If you're pre-PMF, plan one discovery round for the segment you'd bet the quarter on, and keep the pitch out of it. If you're testing an offer, make the ask concrete enough to get a real no. Either way, see how UpSight runs customer discovery with every finding tied to the quotes behind it, or start with the calls you've already recorded.

Try this on three of your own calls.

Free plan, no card. Every answer shows the quote behind it.

Start Free Trial

Founders and product leads: see the customer discovery platform. Consultants: see customer discovery for consultants. Already have recordings? See customer call analysis.