Maybe this is a familiar scene: a founding team has a promising technology, a polished pitch deck, and a product roadmap that grows more ambitious every week. The prototype works. Advisors are encouraging. A few potential customers have said, “Interesting.”
Then comes the difficult part: nobody buys, pilots stall, and the team discovers that “interesting” is not a business model, but often just a polite way of ending a meeting.
Customer discovery matters
Customer discovery means checking whether your key assumptions about the customer, the problem, how buying works, and your value proposition are actually true.
In the NSF I‑Corps program, teams are encouraged to “get out of the building” and talk to potential customers, partners, and other stakeholders to see whether their technology could support a real, sustainable business model.[1]
For deep‑tech startups, this step is especially important because it shapes both what you build and how you plan to launch.
Bare in mind:
Good customer discovery does not validate your idea.
It investigates your risk.
If you treat it as a search for compliments, you will collect compliments.
If you treat it as a search for evidence, you may find a business.
Start with assumptions, not questions
Many founders begin customer discovery by writing an interview script. That is usually one step too late.
Before you ask anyone anything, write down what you believe to be true. For example:
- Who has the problem?
- When does the problem occur?
- How often does it happen?
- What does it cost in time, money, risk, frustration, or missed opportunity?
- How do people solve it today?
- Who would pay for a better solution?
- What would make switching worthwhile?
These are your hypotheses.
And you should test whether they are true or not.
Steve Blank’s customer development approach frames customer discovery as the process of turning a founder’s vision into business model hypotheses and testing customer reactions to those hypotheses.[3]
The practical implication:
your goal is not to “do interviews.”
Your goal is to reduce the uncertainty around the assumptions that could kill the company.
Speak to the right people,
not the easiest people
Early customer discovery goes wrong if founders interview people who are available rather than people who are relevant.
A useful interviewee is someone who
- has direct experience with the problem,
- influences the decision,
- owns the budget,
- uses the current workaround,
- OR feels the operational pain.
In B2B and deep-tech contexts, these may be different people. The user, buyer, technical evaluator, compliance stakeholder, and economic decision-maker may all have different definitions of “value”.
This is why customer discovery should include more than one stakeholder type. So, do not only speak with potential customers but also with partners and other business stakeholders.
For founders, this means that “we talked to ten users” may be useful, but incomplete if no one has spoken to the person who signs the purchase order.
A practical approach is to map the ecosystem:
- Users: Who experiences the problem directly?
- Buyers: Who controls budget?
- Influencers: Who shapes the decision?
- Blockers: Who can say no?
- Current alternatives: What tools, services, spreadsheets, habits, or manual processes are already in place?
Customer discovery does not look for the most enthusiastic voice
but for the most informative pattern.
Ask about behaviour, not opinions
The weakest customer discovery question is also the most tempting:
“Would you use this?”
Most people want to be helpful. Some want to be nice.
A few want the meeting to end before their next call.
Future-intent questions invite all three.
Better questions focus on past and present behaviour:
- “When was the last time this problem occurred?”
- “What did you do?”
- “Who was involved?”
- “What did it cost?”
- “What happens if this is not solved?”
- “What have you already tried?”
- “Did you pay for any of those solutions?”
- “What would need to be true for you to switch?”
This principle is also central to Rob Fitzpatrick’s The Mom Test, a practical customer interview framework that advises founders to ask about the customer’s life and past behaviour rather than pitching the idea or asking for compliments.[4]
While this is an expert practitioner source rather than peer-reviewed research, the advice is consistent with a broader research logic:
Observed behaviour is usually stronger evidence than stated preference.
That does not mean interviews are perfect.
They are not.
People forget, rationalise, exaggerate, and simplify.
Think of the customer interview as a field sensor.
Useful, but occasionally muddy.
That is why strong discovery combines interviews with other evidence:
workflow observation, usage data, landing page tests, letters of intent, paid pilots, procurement conversations, or manual concierge tests.
Listen for pain, workarounds, and consequences
A real problem usually leaves traces.
If a customer has a painful problem, they may already be spending money, time, political capital, or emotional energy on it.
They may have built an ugly spreadsheet, hired someone, delayed a project, accepted regulatory risk, or created a workaround so strange it deserves its own museum label.
In customer discovery, these traces matter.
A useful signal sounds like:
- “We spend two days every month fixing this manually.”
- “We tried three tools, but none worked with our workflow.”
- “If this fails, the project is delayed.”
- “We already have budget for solving this.”
A weak signal sounds like:
- “That is interesting.”
- “Maybe we would use it.”
- “Send me something when it is ready.”
- “I can imagine this being useful for someone.”
The difference is evidence of urgency.
Do not confuse discovery with selling
Customer discovery and sales are related,
but they are not the same activity.
In discovery, your job is to learn.
In sales, your job is to persuade.
Mixing the two too early creates bad data.
If you pitch for twenty minutes and then ask for feedback, you are no longer studying the customer’s problem. You are studying their politeness under social pressure.
That does not mean you should hide what you are building forever.
At the beginning, stay problem-focused.
Once you understand the customer’s world, you can introduce a concept, prototype, or workflow and test reactions.
The goal is to understand what changes, what confuses, what creates value, and what blocks adoption.
A useful structure:
- Context: Understand their role and environment.
- Problem: Explore when and how the problem appears.
- Current behaviour: Learn what they do today.
- Cost: Understand consequences and urgency.
- Decision process: Identify who is involved and what would need to happen.
- Concept test: Only then introduce the solution direction.
- Commitment: Ask for the next concrete step, if appropriate.
The commitment does not always have to be money. In early discovery, a strong signal might be an introduction to a buyer, access to data, a follow-up with the technical team, participation in a pilot, or permission to observe a workflow. The key is that it costs the customer something more than a compliment.
Look for patterns before you pivot
One interview rarely proves anything. Ten vague interviews do not prove much either. The power comes from patterns.
After each conversation, document customer segment, role, problem described, current workaround, perceived cost, urgency, buying process, objections, exact phrases used, next commitment, if any.
Then compare across interviews.
- Are the same problems appearing?
- Are they strongest in one segment?
- Are buyers and users aligned or in conflict?
- Does the problem occur often enough?
- Is the budget real?
- Are people already trying to solve it?
One outcome of customer discovery can be described as substantial first-hand evidence for OR against product-market fit, including clearer customer segments and value propositions.[1]
Good discovery may tell you to continue.
It may also tell you to stop, narrow the segment, change the value proposition, or rethink the route to market.
A pivot discovered through evidence is cheaper than a pivot discovered after twelve months of product development and a website with tasteful gradients.
Common mistakes founders should avoid
The first mistake is asking leading questions: “Wouldn’t it be useful if…?” This turns the interview into a courtroom where the founder is both lawyer and defendant.
The second is interviewing only friendly contacts. Friends, advisors, and innovation managers can be helpful, but they may not represent the buyer with budget pressure.
The third is stopping at problem confirmation. A problem can be real and still not become a business. You also need to understand willingness to change, ability to pay, procurement barriers, and the competitive alternative.
The fourth is outsourcing discovery too early. Agencies can support research, but founders need direct exposure to customer reality. For early-stage teams, the discomfort is part of the data.
The fifth is treating discovery as a one-time phase. Customer discovery is most intense early, but the underlying discipline continues as the market, product, and customer segment evolve.
Conclusion
Customer discovery done well helps you discover
- whether the problem is painful,
- whether the customer is specific,
- whether current alternatives are inadequate,
- whether the buying process is realistic,
- and whether your product direction deserves more investment.
The uncomfortable truth is that good customer discovery may slow you down at the beginning. The useful truth is that it can prevent you from accelerating in the wrong direction.
If you are building a tech-driven startup and want structured support with validation, business model development, market access, and growth, take a closer look at the INiTS SCALEup Incubation Program.
Explore INiTS SCALEup Incubation Program→ inits.at/en/scaleup
The Big Takeaways
Write down your riskiest assumptions before you interview anyone.
If you do not know what you are testing, every conversation will feel useful and very few will be decisive.
Ask about past behaviour, not future interest.
What did you do last time this happened?” is stronger than “Would you use our product?”
Look for costly signals.
A pilot, budget discussion, data access, internal introduction, or workflow observation is more meaningful than “sounds interesting.”
Source list
- U.S. National Science FoundationI-Corps Report2019
https://nsf-gov-resources.nsf.gov/2022-06/I-CorpsReport–6_4_19FINAL_508_0.pdf
- U.S. National Science Foundation2025 NSF Innovation Corps (NSF I-Corps) Biennial Report2025
https://www.nsf.gov/funding/initiatives/i-corps/2025-biennial-report
- Steve BlankCustomer Development for Web Startups2010
https://steveblank.com/2010/02/25/customer-development-for-web-startups
- Rob FitzpatrickThe Mom Test: How to Talk to Customers and Learn If Your Business Is a Good Idea When Everyone Is Lying to You2013