Customer discovery with industrial buyers: nobody has time for your survey

The survey that went to eighty plants and came back empty

An advanced materials startup sent me their discovery results once. Eighty survey invitations to production companies, four responses, all from innovation managers, all politely enthusiastic and all useless. The founder was frustrated: the playbook said customer discovery, they had done customer discovery, and they knew less than when they started.

The playbook is not wrong. It is just written for software buyers who live in their inbox. Your buyer lives on a production floor, measures their day in downtime minutes, and gets approached by twelve vendors a month who all want to "learn about their challenges". Of course they do not respond. Discovery with industrial buyers is the same discipline with a completely different motion.

What customer discovery actually is

Customer discovery is the discipline of proving who has the problem before you build the solution. Steve Blank named it in The Four Steps to the Epiphany and gave it a blunt rule: get out of the building. No opinions from your own team, no survey dashboards, no desk research. Structured conversations with the people who feel the pain, before you try to sell them anything.

In deeptech the definition holds, but the building is harder to leave. Your customer is not a user scrolling an app store. They are a process engineer with twenty years of scar tissue from vendors who promised and vanished. So discovery here is not a phase you finish in a sprint. It is the craft of earning enough trust that a cautious person tells you the truth about their line.

Why the standard discovery motion fails in industry

Three reasons, all structural. First, access: the person with the pain is behind a gate, a procurement firewall, and a calendar owned by someone else. You cannot cold-email your way to a process engineer at scale. Second, language: "what is your biggest pain point?" is a consultant's question, and industrial people have learned to hear it as the opening of a sales pitch, because it always is. Third, stakes: a software buyer who misjudges a tool loses a subscription. A plant manager who misjudges a material or a machine loses production, and possibly their reputation. Their caution is not a discovery obstacle. It is the most important thing to discover.

So the goal shifts. You are not extracting answers to your questions. You are earning the right to hear how their world actually works.

Go where the problem lives

The single best discovery tool in industrial markets is the site visit, and the second best is the trade show hallway. Both put you next to the problem instead of in an inbox. When you stand next to a line that stops twice a shift, you do not need to ask about pain. You can see it, count it, and ask the only question that matters: what happens when this stops?

Getting there requires a different opening. "Can I have thirty minutes to learn about your challenges" gets deleted. "We work on X failure mode in Y process, and before I tell you anything about what we do, I would like to understand how you currently handle it" gets replies, because it leads with their world, not yours. Specificity is the access key. A message that names their process, their likely failure mode, and a plausible cost is read as competence, not spam.

Ask about the last time, not the general case

Once you are talking, the quality of discovery comes down to one habit: ask for the last specific instance, never the general opinion. Not "is downtime a problem?" but "walk me through the last time this line stopped unexpectedly. What time of day was it, who got called, what did it cost, what did you try first?" Specific memories are true. General opinions are performances. The last-instance question also surfaces the buying map for free, because the answer always contains names and roles: who got called, who approved the repair, who complained loudest. Write those down. That is your future stakeholder list.

And listen for the workaround. Every industrial pain has a workaround, some manual inspection, some spare-part ritual, some operator trick, and the workaround is your real competitor. If the workaround costs little and works, your wonderful solution has no market at this account no matter how good the demo looks. Discovering that in week two is a gift.

The questions that work on a plant floor

Rob Fitzpatrick wrote The Mom Test with the right rule: ask about the past, not the future, and never ask anything they can answer with a compliment. On a plant floor I use a fixed set I call the line-stop script. Six questions, in order:

• Walk me through the last time this line stopped unexpectedly. What happened first?
• What did that stop cost, in your numbers? Lost output, overtime, scrap, a late shipment?
• What did you try first, and why that?
• Who got called, and who approved the fix?
• What have you already tried to make it stop happening?
• If you could wave a wand over one part of this process, which part?

Notice what is missing: anything about your product, and anything that starts with "would you". "What keeps you up at night" gets you deleted, because it is a consultant's question and they have heard it twelve times this year. "What happened the last time the line stopped" gets you forty minutes, because it is about their world. The answers also hand you the buying map and the raw material for your first value case, which is exactly how you get to your first ten customers.

How many conversations, and when to stop

Discovery in a narrow industrial segment saturates fast. After ten to fifteen real conversations in one well-defined segment, you will hear the same failure modes, the same constraints, the same names of the two incumbent suppliers. That repetition is the signal. It means the segment's problem structure is mapped and further interviews are comfort, not learning. If you reach conversation fifteen and every story is still different, the news is worse but more valuable: you have not picked a segment yet, you have picked an industry, and industries do not buy. Segments do.

One warning. Five friendly interviews with people who already like you is not discovery, it is reassurance. I see founding teams stop there because it feels like progress. The real test is prediction, not count: you are done when you can write down, before conversation sixteen, what the next plant manager will say, and be right. Until then, keep going.

The output is a one-page problem map

Discovery is finished when you can write, on one page, without adjectives: the failure mode, its frequency, its cost in the customer's own numbers, the current workaround and its cost, who feels the pain, who owns the budget, and what has been tried before. If you can write that page from memory, you have done discovery. If you have a spreadsheet of quotes but cannot write the page, you have done networking.

That page becomes the spine of everything commercial that follows: the pitch, the pilot design, the value case, the pricing. Which is why we treat it as the first deliverable in every engagement, before any outreach begins.

Common questions

What is customer discovery?

Customer discovery is the process of proving who has a problem worth solving before you build or sell. It means structured conversations with the people who feel the pain, not surveys and not pitch meetings. The term comes from Steve Blank's customer development method.

How do you do customer discovery with industrial buyers?

Skip surveys and generic discovery calls. Open with a specific, process-level observation, get to the plant floor or trade show hallway, and ask about the last specific failure instead of general pain points. Ten to fifteen deep conversations in one segment usually saturate the learning.

How many customer discovery interviews do you need?

In one narrow industrial segment, ten to fifteen real conversations are usually enough. You know you are done when you can predict what the next plant manager will say. Five friendly interviews with people who already like you do not count.

Why do industrial buyers not respond to outreach?

Volume and risk. They receive constant vendor approaches, and a wrong decision on the floor costs production, not a subscription. Messages that name their process and failure mode get read as competence and answered; generic pain-point questions get deleted.

What should customer discovery produce?

A one-page problem map: the failure mode, frequency, cost in the customer's numbers, current workaround, the person who feels the pain, the budget owner, and what has been tried. If you cannot write it from memory, keep going.

If you have done the interviews and still cannot write the one-page problem map, that is exactly where we start. Tell us what you have built.