Customer discovery in MedTech is not one activity. It is three activities that share a name and nothing else. You interview clinicians to understand the workflow the device has to survive. You interview patients to understand the lived experience of the condition the device is supposed to improve. You interview hospital buyers to understand the economics of the purchase order that will eventually be signed. Three different people, three different question sets, three different signals, three different definitions of a useful answer. A founder who runs one generic "talk to users" script across all three learns nothing from any of them and usually walks away convinced they have validated the idea because everyone was polite. Run three separate discovery tracks, with three different scripts, and treat every conversation as a verification of the market rather than a pitch of the product.

By Felix Lenhard and Tibor Zechmeister. Last updated 10 April 2026.

TL;DR

  • MedTech customer discovery is three tracks, not one. Clinicians, patients, and hospital buyers each need their own script, their own sample, and their own definition of signal.
  • Clinician interviews are about workflow and friction. Patient interviews are about lived experience with the condition. Buyer interviews are about economics and the purchase order.
  • You cannot run a pre-CE clinical trial disguised as a user interview. Interviewing patients about their condition is fine. Putting an uncertified device in a patient's hand to "see how they like it" crosses into territory the regulation governs under Article 62 and following.
  • Signal in customer discovery is not "they liked it." Signal is a specific pain named in specific language, a quantified cost, a willingness to change workflow, or a letter of intent.
  • Document every interview the day you run it. Investors want to see the pattern across twenty conversations. Regulators want to see the evidence base behind the intended purpose sentence.
  • A founder who runs customer discovery well arrives at the regulatory project with a sharp intended purpose, a defendable economic model, and three to five clinical partners already on side. A founder who skips it arrives with a prototype and a theory.

A EUR 1.8 million lab product that never met a real user

A founder I coached spent EUR 1.8 million on a device that worked in a lab and failed in the real world. That is the short version. The long version is more painful, and it is the reason this post exists.

The device was elegant. The engineering was clean. Every demo inside the company's own building ran exactly as designed. The team had talked to clinicians — sort of. They had shown the device at conferences. They had taken meetings with heads of department. They had given presentations and collected polite nods. What they had not done was sit in a ward for a day and watch the exact people who would eventually use the device do the exact work the device was supposed to change. They had not asked a patient what the condition actually felt like from inside their own body. They had not sat across from a hospital procurement director and asked what it would take to put this device on a purchase order.

They had run the shape of customer discovery. They had not run customer discovery.

When the device finally left the lab and met real users in real settings, the users did what users always do. They skipped steps. They held it wrong. They ran it at three in the afternoon in a noisy ward with a trainee nurse they had never worked with before. The device did not break. The device worked perfectly. The assumptions the founders had built it on were the thing that broke. Eighteen months of work, one point eight million euros, and a product nobody could use. That is what happens when you confuse talking about your idea with discovering what your customers actually need.

Now compare that with a different founder I coached in an adjacent domain. He did something that looks almost absurdly simple. He built a landing page, had conversations with buyers, sold the offer, and only then built the product. EUR 50,000 of revenue in the first quarter on zero marketing spend. People were actively asking him to take their money. The reason it worked is that he had done real discovery before he had anything to sell. He had answered the one question every buyer had, and he had charged money for the answer.

You cannot run that exact playbook in MedTech because you cannot sell an uncertified device to patients. But you can run the discovery half of it rigorously, and you can do it before you spend a single euro on the regulatory project. This post is about how.

Why MedTech customer discovery is different

Customer discovery as a concept comes out of the lean startup world. Talk to users, learn what they want, iterate. That works in SaaS because the user is the buyer is the beneficiary — one person, one wallet, one workflow. In MedTech none of that alignment holds.

Three things make MedTech customer discovery a different discipline.

The customer is not one person. In a hospital or clinic setting, the clinician uses the device, the patient benefits from the device, and the hospital pays for the device. These are three different humans with three different pains and three different languages for value. A SaaS-style "talk to your users" approach picks one of those three, usually the clinician, and mistakes the signal for PMF. It is not. A clinician's enthusiasm is necessary and insufficient. You need all three.

You cannot iterate with patients the way SaaS iterates with users. Under Article 2(1) of Regulation (EU) 2017/745, a medical device is defined by the medical purpose the manufacturer claims for it, and the moment you put an uncertified device into a patient's hand for that purpose you have crossed a legal line. (Regulation (EU) 2017/745, Article 2, paragraph 1.) The regulation lays out a specific path for testing devices that are not yet certified — clinical investigations under Article 62 and following. (Regulation (EU) 2017/745, Article 62.) That path exists and it is the right tool when the time comes. It is not a tool for casual discovery.

The buyer has a completely different pain than the user. A clinician who loves your device will tell you about minutes saved per procedure, friction removed from a handover, a complication she no longer worries about. A hospital budget holder in the same building will tell you about cost per case, throughput per day, readmission rates, and reimbursement codes. Those are not the same language. A founder who only runs the clinician script will never discover the budget holder's objection until they have already built the wrong device.

The practical consequence is that MedTech customer discovery is three tracks. You do not run one script across all three audiences. You run three scripts, in parallel, with three samples, and you treat the intersection of the three as the place where PMF lives.

What to ask clinicians versus patients versus buyers

Each track has a purpose, a question shape, and a definition of signal. Here is how to think about each of the three.

Clinician interviews exist to understand the current workflow and the friction your device would change. You are not pitching. You are learning. The best clinician interviews feel like the founder is apprenticing for an afternoon. The questions are about what happens right now, not about what the clinician thinks of the idea. Useful clinician questions sound like these. Walk me through the last time you did this procedure — what happened, in order, minute by minute. What goes wrong with the current approach, and how often. What do you do when it goes wrong. How long does the task take on a good day and on a bad day. Who else is in the room when you do it. What do you wish were different, and why has nobody fixed it yet. Every answer is a piece of the workflow map. Signal is a specific, vivid pain described in the clinician's own words, repeated across multiple interviews.

Patient interviews exist to understand the lived experience of the condition — not the device, the condition. You are learning what it is actually like to live with the problem your device is supposed to solve. Patient questions sound like these. When did you first notice the condition. What does a typical day look like now. What is the hardest part of it. What have you tried, what worked, what did not, and how do you know. When you go to your appointments, what happens there, and what do you wish were different. How has it affected work, family, sleep, mood, cost. Notice that none of these questions involve the device. You are not asking the patient to evaluate your product. You are asking the patient to describe the condition that gives your product a reason to exist. The signal is a coherent, specific picture of the condition that your clinical evaluation will later have to engage with honestly.

Hospital buyer interviews exist to understand the economics of the eventual purchase order. The buyer is the head of procurement, the medical director, the CFO, the head of nursing, the managing director of a private clinic, or in some systems a value analysis committee. The buyer questions are about their budget, not about your device. How is this class of spend funded today. What does it take to add a new supplier. What is the approval chain for a device in this category. What size of purchase moves through which committee. When was the last time a device in this area was purchased, and what drove the decision. What would have to be true for you to put this device on next year's budget. What would make you say no in the first five minutes. The signal is a clear economic story — a named pain, a quantified cost, and a plausible path from the pain to a purchase order. If you cannot draw that path after twenty buyer conversations, the buyer economics are not there yet.

Three tracks, three question sets, three definitions of signal. Never merge them into a single generic script.

The ethics of patient interviews pre-CE

There is a legitimate question every responsible MedTech founder should ask early. Is it ethical to interview patients about a device that does not yet exist, before the regulation has had a chance to say anything about it? The short answer is yes, as long as you understand the line.

The line is between talking to a patient about their condition and putting an uncertified device into a patient's hand as if it were a medical product. The first is ordinary market research. Journalists do it, researchers do it, founders can do it. The second is regulated territory. Under Article 62 and following of Regulation (EU) 2017/745, the use of a device on human subjects for the purpose of generating clinical evidence falls inside the clinical investigation framework, with all the documentation, ethics, and oversight that entails. (Regulation (EU) 2017/745, Article 62.) That framework exists for good reasons, and it is the right path when the time comes.

For early customer discovery, the guardrails are simple. Talk to patients about their condition, their journey, their current treatment, and their lived experience. Do not hand them an uncertified device. Do not let the conversation drift into "would you try this next week." Do not run an unmonitored evaluation of a prototype in a patient's home under the banner of user research. If your discovery needs any hands-on use of the device by a patient, that is no longer discovery. That is an investigation, and it belongs inside the Article 62 framework, with proper clinical partners and proper oversight.

There are two more boundaries to respect. First, informed consent for the interview itself — patients should know what the conversation is for, what will be written down, and how their story will be used. Second, data protection — patient information is sensitive under GDPR, and founders should not build informal databases of identifiable patient stories on personal laptops. Both of these are cheap to get right if you plan them into the discovery process from the start.

Patient interviews done inside these guardrails are among the most valuable conversations a MedTech founder can have. They are also among the least commonly done, because they are harder to schedule than a clinician call and the founder cannot control the agenda the way they can control a buyer meeting. The founders who do them well arrive at the regulatory project with a story about the patient that a clinical evaluator would actually believe.

How to recruit for each track

Each of the three tracks has a different recruiting problem.

Clinicians are recruited through their own networks. A cold email to a senior physician almost never works. A warm introduction from a mutual contact works well. Conferences, specialty societies, LinkedIn, alumni networks, and your existing advisory board are all legitimate entry points. The rule is to respect their time. Fifteen to thirty minutes per interview, a specific question list sent in advance, and an offer of something in return — usually a summary of what you are learning across the cohort. Never dress up a sales call as an interview.

Patients are recruited through patient associations, support groups, clinician referrals (with proper consent), and in some cases through dedicated market research panels. The recruiting has to be transparent — who you are, what the research is for, how long it will take, what happens to their story. For rare conditions, a single well-run patient association partnership can unlock interviews that would otherwise take months. For common conditions, clinician referrals work well because the clinician is already acting as a trusted gatekeeper.

Hospital buyers are the hardest to recruit and the most valuable per interview. They are recruited through the clinician network ("can you introduce me to your head of procurement"), through industry associations, through your own advisory board, and through the commercial contacts of any existing suppliers you can find. Buyers are not usually impressed by founder enthusiasm. They are impressed by specificity — a clear description of what you are asking, why you are asking them in particular, and a promise to use their time efficiently. Twenty minutes of a buyer's time is worth an hour of almost anyone else's, and the founder has to show up ready to use it well.

A reasonable discovery cohort for a MedTech startup entering a new segment is in the range of twenty clinician interviews, ten to twenty patient interviews where the condition permits, and twenty buyer interviews. Fewer than that and you are pattern-matching on noise. More is fine. The numbers are not magic — they are the point at which the objections stop surprising you.

What counts as signal

The most common mistake in customer discovery is mistaking politeness for validation. Clinicians are trained to be constructive. Patients are usually grateful for the attention. Buyers are diplomatic because their reputation is on the line. None of those three will tell a founder their idea is bad in the first meeting. Founders who do not know what to listen for walk out of every meeting convinced they have validation.

Signal is specific, not general. Here is what to listen for, per track.

From a clinician, signal is a specific pain named in the clinician's own words, without prompting, and a concrete description of what the clinician does today to work around the pain. If the clinician volunteers a number — "I waste about forty minutes every shift on this" — that is stronger signal. If the clinician mentions the pain without being asked about it, that is stronger still. "Would you use a device that did this?" is not a question that produces signal. "What do you do right now when X happens?" is.

From a patient, signal is a coherent, specific description of the condition and the current care pathway, including the parts that do not work. Strong signal includes named costs — time, money, relationships, quality of life — and specific moments when the condition got in the way of something the patient cared about. Weak signal is agreement with leading questions. A patient who says yes to "would this help you" is being polite. A patient who describes a specific morning when the condition cost her something real is giving you gold.

From a buyer, signal is economic. Is there a line item in the budget that this device would affect. Is there a funding mechanism that would cover it. Has the buyer put a device like this on the budget in recent memory, and what drove the decision. Would the buyer put this on next year's budget, and under what conditions. A letter of intent from a buyer is harder signal than any verbal enthusiasm. A buyer who walks you through the approval chain without being asked is showing you the path to a real purchase order.

Signal is also what is not said. If a clinician hesitates on the workflow question, if a patient steers the conversation away from a topic, if a buyer cannot name the budget line, those are signals too. The founders who run discovery well listen as carefully to the hesitations as to the words.

How to document findings for investors and regulators

Every discovery conversation has to be written down the day it happens. Memory decays fast, and by interview number ten the pattern only exists on paper. The founders who do this well build a simple structure that serves two audiences at once — investors and regulators.

For each interview, record the role and context of the person, the date, the questions asked, the verbatim answers that matter, and the founder's own interpretation of what the conversation means. Keep the verbatim separate from the interpretation — this is a discipline that pays back later when you are tempted to read patterns that are not there. Tag each interview with the track it belongs to (clinician, patient, buyer) and the hypothesis it tests or breaks.

At the end of a discovery phase, produce three short summaries. A clinician workflow map describing the actual current state, the friction points, and the quantified pains. A patient journey map describing the lived experience of the condition and the costs of the current care pathway. A buyer economic model describing the funding path, the approval chain, the comparable purchases, and the conditions under which a purchase order would be signed.

Those three documents are what investors want to see when they ask whether you have PMF. They are also the evidence base behind the intended purpose sentence you will later write under Article 2(12) of the MDR. (Regulation (EU) 2017/745, Article 2, paragraph 12.) An intended purpose written on top of real discovery documentation is specific, defendable, and classifiable. An intended purpose written from imagination is vague, shifts under pressure, and triggers expensive rework later. The documentation work you do during discovery is not separate from the regulatory work. It is the foundation of the regulatory work.

Common mistakes

  • Running one generic script across all three audiences. A question set built for clinicians is the wrong shape for patients and the wrong shape for buyers. Build three scripts.
  • Pitching instead of listening. If the founder is talking more than the interviewee, it is not discovery. It is a demo the interviewee is too polite to interrupt.
  • Interviewing only champions. A cohort made up entirely of enthusiastic clinicians tells you nothing about the skeptical clinicians who will outnumber them in any real deployment. Deliberately recruit people who are not already fans.
  • Skipping the buyer track. The most common single mistake. Founders love clinician interviews because they feel productive. Buyer interviews are harder, slower, and more uncomfortable. They are also the ones that produce the signal that saves the regulatory budget.
  • Crossing the line with patients. Handing an uncertified device to a patient to "see what they think" is not research. It is an unmonitored use of a medical device, and it belongs inside the clinical investigation framework, not inside a discovery sprint.
  • Not documenting the day of. By week three, the interviews blur. The pattern only exists if the notes exist.
  • Declaring validation after ten interviews. Twenty per track is the floor, not the aspiration. Stopping early is how founders validate their own assumptions instead of the market.
  • Confusing enthusiasm with economic commitment. "This looks interesting" is not signal. "I could see this on our budget next year if X were true" is signal.

The Subtract to Ship angle

Subtract to Ship applied to customer discovery is a short list. Every conversation should belong to one of the three tracks — clinician, patient, or buyer — and every conversation should have a pre-written question list aimed at that track. Conversations that do not fit the tracks are usually founder therapy, not discovery, and they should be cut. Every interview should produce documentation the day it happens, in a format that will feed both the investor pitch and the intended purpose sentence. Anything else is waste.

Read how to validate a MedTech idea before MDR for the full eight-step pre-regulatory validation sequence this post sits inside, and the Subtract to Ship framework for MDR for the same discipline applied across the whole certification project.

Reality Check — Where do you stand on discovery?

Answer these honestly. If more than two are weak, your discovery is not done and you are not ready to commit to a regulatory budget.

  1. Do you have three different interview scripts — one for clinicians, one for patients, one for buyers — or a single generic one you use for all three?
  2. How many clinician interviews have you run in the last ninety days, and can you describe the workflow pain they identified in the clinicians' own words?
  3. How many patient interviews have you run in the last ninety days, and can you describe the lived experience of the condition in a way a clinical evaluator would recognise?
  4. How many hospital buyer interviews have you run in the last ninety days, and can you name the budget line and the approval chain a purchase order would travel through?
  5. Have you recruited any skeptical interviewees on purpose, or is your cohort entirely people who already like the idea?
  6. Have you drawn the line between patient discovery interviews and unmonitored device use with patients, and have you stayed on the research side of that line?
  7. Do you have written notes from every interview in a format that separates verbatim from interpretation?
  8. Can you produce a one-page clinician workflow map, a one-page patient journey map, and a one-page buyer economic model that together justify the intended purpose sentence you are about to write?
  9. Has a clinician who is not on your cap table read your intended purpose sentence and said it matches what they would sign off on?
  10. Has a hospital buyer who is not your cousin said, in writing, that they would consider this device under stated conditions?

Frequently Asked Questions

What is MedTech customer discovery? MedTech customer discovery is the pre-regulatory practice of interviewing clinicians, patients, and hospital buyers separately to verify the market for a medical device before a regulatory budget is committed. It is three parallel tracks, not one generic "talk to users" activity. Clinician interviews map the workflow. Patient interviews map the lived experience of the condition. Buyer interviews map the economics of the purchase order. PMF lives at the intersection of all three.

How is MedTech customer discovery different from SaaS customer discovery? In SaaS the user, the buyer, and the beneficiary are usually the same person, so a single script works. In MedTech they are three different people with three different pains and three different definitions of value. You cannot iterate an uncertified device in a patient's hand the way a SaaS founder iterates a signup flow, because the regulation draws a line around that kind of use. Discovery in MedTech therefore has to be three tracks run in parallel, with three different scripts and three different definitions of signal.

Is it legal to interview patients about a medical device that does not exist yet? Yes, when the conversation is about the patient's condition, care pathway, and lived experience, not about hands-on use of an uncertified device. That kind of interview is ordinary market research and is governed by the usual informed consent and data protection rules, not by Article 62 of the MDR. The moment the conversation crosses into the patient actually using an uncertified device for its intended medical purpose, you are inside the clinical investigation framework and different rules apply. See pre-CE clinical investigations for the detail on where that line falls.

How many interviews do I need before I can say I have validated the market? Around twenty per track — twenty clinician interviews, twenty buyer interviews, and as many patient interviews as the condition and recruiting permit. Fewer than that and you are drawing conclusions from too small a sample. More than that is usually fine. The real stopping rule is that the objections stop surprising you. If every new interview is still showing you a pain you had not heard before, keep going.

Who should do the customer discovery — the founder or a hired researcher? The founder. Discovery is not a task to delegate. The pattern you are trying to see only emerges if the same person sits through all the interviews and carries the developing picture from one conversation to the next. A hired researcher can assist with recruiting, documentation, and scheduling. The founder has to be in the room.

What do I do if clinicians love the device and buyers do not? You do not have PMF yet. Clinician enthusiasm without buyer economics is the single most common pattern in failed MedTech companies. The fix is usually one of three things — find a different buyer whose economics do align, reposition the device so its benefit maps to something the existing buyer actually counts, or change the device so it moves a metric the buyer cares about. See decision-making units in MedTech sales for the full treatment of the champion-versus-buyer trap.

When does customer discovery end and formal clinical investigation begin? Discovery ends when the intended purpose sentence is stable, the three summary documents are in place, and the founder is ready to commit to a regulatory path. Clinical investigation under Article 62 and following of the MDR begins when the device is far enough along to be tested on human subjects in a controlled clinical setting, with full documentation, ethics approval, and clinical partners. The two phases do not overlap. Discovery verifies that the device is worth building. Clinical investigation generates the evidence that it works.

Sources

  1. Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices, consolidated text. Articles cited: Article 2(1) definition of medical device; Article 2(12) intended purpose; Article 62 clinical investigations. Official Journal L 117, 5.5.2017.
  2. Regulation (EU) 2023/607 of the European Parliament and of the Council of 15 March 2023 amending Regulations (EU) 2017/745 and (EU) 2017/746 as regards the transitional provisions for certain medical devices and in vitro diagnostic medical devices. Official Journal L 80, 20.3.2023.
  3. Felix Lenhard coaching case files, 2021–2026, anonymised for publication.

This post is part of the MedTech Startup Strategy and PMF cluster in the Subtract to Ship: MDR blog. Authored by Felix Lenhard and Tibor Zechmeister. Run three discovery tracks, not one. Listen harder than you pitch. Write it down the day it happens. Your intended purpose sentence is only as sharp as the conversations behind it.