Walk through the US Patent Office archives and you find a museum. Thousands of patents for things built beautifully, engineered rigorously, solving a problem nobody had. Each represents years of somebody’s life.
Mark DeSantis has a name for the object at the center of that museum. He calls it the platinum hammer: a tool of genuine craftsmanship, extraordinary in its construction, looking for a nail that failed to exist.
In Episode 16 of Growth Files, Sathish Kumar speaks with Mark DeSantis, a five-time founder whose companies have been acquired by Michelin and Kubota, and who teaches entrepreneurship and AI at Carnegie Mellon. The conversation covers why technologists fall for their own solutions, what a hundred customer interviews reveal, and how to tell polite encouragement from a real buyer.
Episode TL;DR
- •Science discovers, engineering builds, business solves problems people pay for. Those three goals point in different directions.
- •Get a PhD in the problem. An associate’s degree in the solution will do.
- •Customer interviews produce encouragement, and encouragement gets misread as demand.
- •Go looking for evidence that your problem is imaginary. That is the only honest test.
- •Talk to a hundred people about the problem and mention your solution to none of them.
- •Everything before product market fit is a project. Everything after it is a business.
- •Surveys answer narrow questions well and stay silent on latent needs.
- •AI expands the breadth of your thinking. Humans still hold the depth.
- •Nothing in entrepreneurship is failure. It is all learning, and it is expensive learning.
Table of Contents
- 1.The Platinum Hammer Problem
- 2.Why Technologists Fall in Love With the Solution
- 3.The Swiffer Test: Why Real Problems Take 150 Years to Find
- 4.Needs, Wants and Why Passion Decides
- 5.PhD in the Problem, Associate’s Degree in the Solution
- 6.Why Customer Interviews Lie to You
- 7.Prove the Null Hypothesis
- 8.Co-Creation and How Product Market Fit Actually Happens
- 9.Everything Before Product Market Fit Is a Project
- 10.The Hundred Interviews
- 11.Can AI Replace Customer Discovery?
- 12.What Yes Sounds Like, and Why Nothing Here Is Failure
About the Guests
Guest
Mark DeSantis, five-time founder and adjunct professor of entrepreneurship at Carnegie Mellon University, with appointments in both Heinz College and the College of Engineering. Most recently CEO of Bloomfield Robotics, acquired by Kubota. Previously cofounder and CEO of RoadBotics, acquired by Michelin, cofounder of kWantix and kWantera, and CEO of Think Through Learning. Earlier roles include director of government relations at Texas Instruments and senior policy analyst in the White House Office of Science and Technology Policy. PhD in public policy from George Mason University. He writes Real AI = Real Work on AI that does actual work.
Host
Sathish Kumar Mariappan, CEO and Co-Founder of CommerceShop and host of Growth Files. Sathish leads a revenue-first eCommerce consultancy for brands scaling from $2M to $25M, focused on conversion optimization, answer engine optimization and B2B manufacturing commerce.
The Platinum Hammer Problem
Twenty years of building companies produced one finding worth leading with, and it concerns what fails. The pattern shows up most reliably in companies built by technical founders, and it starts with a mismatch of purpose.
Mark: “Scientists like to discover things, and engineers like to build things. And the problem when engineers and scientists become business people is that business doesn’t care about discovery, nor does it care about building for the sake of building. What it cares about is solving a problem.”
Three disciplines, three goals, no guarantee they point the same way. Science pushes the boundary of knowledge. Engineering advances what can be built. Business solves a problem somebody will pay to have solved. A founder carrying the first two instincts into the third brings the wrong reflexes.
What goes missing is that whatever gets built has to serve a clear human need, and in most startups he has seen, that happens by accident. The sequence repeats at every university spinout pitch. Five years of research yields something genuinely impressive. It must have an application somewhere. Raise money, build it, find out whether anyone buys.
Mark: “That’s a hammer. Platinum. Platinum hammer. Well built, extraordinary, very interesting. And it doesn’t solve any problem or satisfy any human need that anyone is willing to pay for.”
The word platinum carries the metaphor. A cheap hammer lacking a nail is a small waste. A platinum one represents years of scarce talent, serious capital and real craftsmanship, all aimed at nothing.
Why Technologists Fall in Love With the Solution
If the trap is this well known, why do founders keep walking into it? The answer starts by defending the researchers.
Scientists exist to push the boundary of human knowledge, and the discovery is the reward. Plenty stayed poor and changed the world anyway. Academic research aimed at commercial return would be worse research and deserves to be left alone. The problem sits elsewhere.
Mark: “There are people who do not understand the academic research or the research of scientists and engineers, see it as a goldmine of possibilities and throw money at it.”
At Carnegie Mellon this happens in AI specifically, where researchers are offered a hundred million dollars or more to leave their post and turn a research program into a product. Who could resist?
Money introduces the distortion. A large check reads as external validation, and founders reason backwards from it.
Mark: “People are convinced that if I’ve got a hundred million dollars, I must be on to something. And the answer is no. That’s not true at all.”
The survivorship problem becomes concrete in the pet aisle. Every dog food brand on that shelf is a survivor, and the hundreds that failed left no trace, which is what makes the shelf misleading as evidence.
Mark: “This is Broadway for dog food. If you’ve made it onto the shelf of Safeway or Whole Foods, you’ve made it on Broadway. And there are hundreds of thousands of actors that never made it to Broadway.”
Tech works the same way. Funding announcements are visible and quiet failures are silent, so a founder calibrating against what they can see will overestimate the odds.
The Swiffer Test: Why Real Problems Take 150 Years to Find
Technical founders tend to hear identifying human needs as soft work next to research. It is a discipline in its own right, and harder than it looks.
Mark: “Nobody these researchers talk to is gonna say, thanks for coming to my office, let me tell you about my needs and wants, and if you build this I will buy 10,000. Said no one in the history of the world, and meant it.”
The illustration is a mop. The CEO of Procter & Gamble spent weeks in Mexico inside the home of a middle-class family, watching them clean.
Mark: “What he saw was that mops don’t work. Mops may work on the deck of a sailing vessel 300 years ago. But cleaning your bathroom with a mop, not so much. They just push dirt around, then you squeeze them off in this bucket of dirty water.”
That observation produced the Swiffer. The timeline explains everything.
Mark: “Procter and Gamble has been around making soap and detergent since the 1850s. It took 150 years for them to develop the Swiffer. What does the Swiffer do? Cleans your floor.”
A century and a half, inside one of the largest consumer goods companies in the world, dedicated to this exact category, before anyone improved on a mop. That is the real measure of how hard latent need discovery is.
Mark: “It’s hard, but once you’ve done it, it looks easy, and that’s the danger.”
A solved problem looks trivial in hindsight, which leads founders to underestimate the work of finding the next one.
Needs, Wants and Why Passion Decides
Sathish raises the distinction that decides whether a business exists: people say they need something and then decline to pay. Both sides turn out to be latent.
Mark: “I want a McLaren car. But do I need to spend four million dollars on a car? No, I don’t.”
What a founder hunts for is a latent need, and the variable that predicts whether they find it is unexpected.
Mark: “If you don’t have a passion for that problem, it’s hard to turn that into a successful business. I’ve done it myself. I convinced myself I was passionate about it. Not sure I was.”
The mechanism is attention. Latent needs live in detail, and anyone who finds the subject dull misses the detail.
The example is Steve Jobs, framed in terms most Jobs commentary skips. The question was how a person interacts with a screen, and the skill was sustained attention to something apparently mundane.
No training in it, no technical edge over the people working for him. The difference was passion for the question. An Apple device carries few buttons and dials, and that cleanliness was engineered for accessibility, well beyond aesthetics. Getting there took a volume of thought that only passion sustains.
Hence the diagnosis.
Mark: “They’re passionate about the solution. They’re passionate about their platinum hammer. They’re not passionate about the problem it’s going to solve.”
The prescription follows: find the passion that motivated the original research, then translate it into a real problem. The energy already exists, aimed at the wrong object.
Watch the Full Conversation on YouTube
Mark walks through the hundred-interview method, the five whys applied to customer discovery, and the vineyard meeting that told him he finally had a business.
Watch on YouTube →
PhD in the Problem, Associate’s Degree in the Solution
The line he gives his Carnegie Mellon students compresses the whole argument.
Mark: “I say PhD in the problem, and you can get away with an associate’s degree in the solution.”
The claim underneath is uncomfortable for technical teams. Winning solutions are frequently unremarkable as engineering.
Winning products are frequently unremarkable engineering, at least to other engineers.
One tell shows up repeatedly. Researchers examine an incumbent product, find it technically unsophisticated, and conclude they can do better. What they mean by better is internal: faster throughput, improved mean time between failures, a percentage gain somewhere in the architecture.
Mark: “What they don’t appreciate is that product was built for a very specific need that took a lot of work to understand, and these people don’t even see that problem. They’re hung up on the solution.”
The consequence is a specific disappointment, and it hits even teams that succeed. Founders who assemble a Nobel laureate advisory board and serious technology often find the thing that finally sells strikes them as mundane.
Founders rarely expect that disillusionment. The product that finally sells is often the least technically interesting thing the team built, and anyone measuring satisfaction by technical elegance will read commercial success as anticlimax.
Why Customer Interviews Lie to You
Sathish puts the practical objection: in interviews people say they need the thing and would buy it, then do neither. The explanation is social.
Picture the scene from the other side of the table. Three Stanford graduate students call an executive to discuss his company’s problems, and the meeting gets booked.
Mark: “I’m gonna cheer them on. Am I gonna buy what they’re gonna make? No, absolutely not. But they’re motivated and what they’re doing is cool, and I don’t want to discourage them. I was an entrepreneur once, says the president of this company.”
Everyone was sincere. The executive was generous, the founders were earnest, the meeting went well. Then the founders read encouragement as a signal they are onto something, when what happened was a pleasant conversation with a thoughtful person who wanted them to succeed.
Interviews with a solution in the room generate warmth. Warmth is indistinguishable from interest when you want it to be, and a founder collecting encouragement reads a full calendar as validated demand.
Prove the Null Hypothesis
The fix borrows a frame technical founders already respect, which is why it lands.
Mark: “Your goal in finding an opportunity, a problem or need, is you’re actually trying to prove the null hypothesis. You’re basically trying to prove that your belief that there’s a problem is in fact not true. You’re actually going out there and finding evidence counter to the idea that this problem or need exists.”
The reason to invert the test is confirmation bias, a human constant more than a personal failing.
Mark: “I’m passionate about starting a company. I so badly want to start a company. So what do I do as a human being? I go out and find data that confirms my bias. Has that ever happened in the history of the world? Even among well educated people?”
Volume alone provides no protection. Six people who think your work is terrific and sixty people who think your work is terrific tell you the same thing, which amounts to zero information.
The alternative changes one variable and everything follows from it.
Mark: “That’s different than I’ve talked to 60 people about a problem and did not once mention our solution. I had 60 conversations where I never once mentioned our agentic AI. You know what I learned? Forty of them said this is not a problem.”
Forty rejections out of sixty counts as the successful outcome. A conversation surfacing indifference has given you information. A conversation full of enthusiasm about your solution has given you a mood. The discipline is holding the solution back long enough for the honest answer to arrive. Sathish notes he has hit exactly this in his own interviews, which is the common experience for anyone running discovery while wanting a particular answer.
Co-Creation and How Product Market Fit Actually Happens
A good interview is neither extraction nor pitching.
Mark: “You and the strangers you’re meeting are going on a journey together of co-creation. You both start in different places and you’re not moving toward each other. You’re moving together in the general direction of something else.”
The mechanism is a dialectic. What about this, what about that, with each exchange refining your understanding of their needs and their understanding of what is now possible. Where it lands has a name everyone already uses, and it comes with a warning attached: you eventually reach product market fit, and that is when the real headaches begin.
What separates this from conventional customer research is that the customer learns too. They come in with a partial view of what is now possible. Product market fit becomes a destination neither side could have specified at the start, which is why walking in with a fixed spec and a question list yields less from the same hour.
Everything Before Product Market Fit Is a Project
A caution comes before the framework, and it is rare enough to repeat.
Mark: “How did I learn what I’ve learned? Because I’ve made all the mistakes. I could probably come up with some kind of patent of how to make mistakes.”
A second caveat concerns generalizability. That journey was specific to a time and place, every founder’s differs, and what follows is experience corroborated by other entrepreneurs.
The definition that follows is the cleanest in the episode.
Mark: “Everything before product market fit is a project. Everything after product market fit is a business. A going concern is: I make a thing that you buy, and ideally you buy it for more than it costs me to give it to you.”
Then the Drucker line, with a correction attached.
Drucker’s line that the purpose of a business is to create a customer sits underneath this, with one word doing the work. Founders say they want to find a customer, and finding assumes the customer already exists.
Mark: “You have to create a customer. There was no pre-existing market of people that said, boy, if I just had a phone that didn’t have a cord that I could put in my pocket, then I would buy it. It’s obvious now why wouldn’t you want a cell phone. But it wasn’t obvious then.”
Sathish raises the tension every founder has heard: Ford’s faster horse, and Jobs on the limits of asking customers. Both stories get used to justify skipping discovery, which is the wrong lesson. Neither man skipped understanding the problem. What they declined to do was ask customers to specify the solution.
Listen to the Full Episode on Spotify
The audio version covers the Deming five whys applied to interviews, why surveys fail at discovery, and Mark’s take on effectiveness over efficiency in AI.
Listen on Spotify →The Hundred Interviews
An opportunity legitimately originates in personal experience of the problem.
Mark: “Your belief that there’s a problem that exists in the world that you yourself have experienced. Just because you’ve seen it doesn’t mean a lot of other people do, but it does mean some people do.”
From that ember comes a question posed at deliberate scale.
Mark: “Eight billion people in the world. Who are the hundred most likely to feel this need most acutely? And then I go talk to them.”
Selecting those hundred is instinct more than method. The worked example is a math teacher struggling to hold teenagers’ attention through geometry and trigonometry. One rule governs those conversations, and it holds without exception.
Mark: “I’m going to interview a hundred math teachers who do not know me, who don’t care if I win or lose. They won’t hear AI come out of my mouth, solution come out of my mouth, product. All we will talk about is the challenge of teaching trigonometry to teenagers.”
The interviews change the founder as much as the plan.
A hundred of those conversations changes the founder as much as the plan. You come out knowing the problem better and knowing yourself better, including whether you actually want to do this. Steve Blank at Stanford pushed the practice under the phrase get out of the building, and one legitimate outcome is a decision to stop. People have completed the hundred interviews and concluded they would rather stay a professor or keep the job at Google, which counts as the process working.
On surveys. The efficient-looking shortcut fails here for a specific reason.
Mark: “Surveys are great if you want to know who you’re going to vote for in the upcoming election, or if you have a very specific piece of information you’re seeking from a very specific population. That’s not what you’re doing.”
Discovery needs the channel that carries nuance: body language, tone, the random digression. Understanding a latent need is a human effort, and the tools belong to social science more than statistics.
On the shape of the hundred. The interviews do not improve steadily, and the early chaos is what makes founders quit.
Mark: “You may find that in the first 30 or 40 interviews you’re asking the wrong question and talking to the wrong people. That’s where people get discouraged. And I say, quite the contrary, that’s the good news. You’re trying.”
The arc runs in thirds. The first reveals that both question and audience are wrong. The second corrects the question and narrows the audience. By the last twenty or thirty conversations you have found the people who genuinely hold the problem, and that is when you go deep.
On the questions themselves. The tool comes from total quality management, developed in postwar Japan under W. Edwards Deming: the five whys, chasing a defect down through layers until the root appears. Point it at a person. Why are students confused by trigonometry? Because parts of it are unintuitive. Why? Because there is no way to visually present what happens with geometric forms. Why? And onward, until the problem appears at its elemental level.
Mark: “That’s co-creation. You’re facilitating thinking by that person that they’ve never done themselves. It’s not a deposition, as lawyers say.”
A deposition extracts facts the other person already holds. A five whys conversation builds understanding neither party walked in with, and the moment both people reach something new is the signal the hour paid off.
Can AI Replace Customer Discovery?
Sathish asks the obvious question: with LLMs available, can a founder run the five whys against an AI and skip some of this?
Mark: “Potentially, but I think it can create the illusion of that. As complicated as LLMs are, humans are more complicated. And that’s where the information still resides.”
AI is a daily tool for him, and the axis it helps on is specific: it expands the breadth of your thinking, and it struggles to give you the depth a human conversation does. The synthetic alternative gets entertained openly, a billion simulated humans available for interview, left as an invitation to whoever wants to build it. On today’s capability the position stays firm.
Mark: “The best source of information about human wants and needs and fears still comes from humans. The machine can’t provide that yet.”
AI is a strong instrument for generating question sets, pressure-testing a hypothesis and preparing before a conversation. The latent need itself sits in nuance a person may not have articulated even to themselves, which is precisely what a model has no access to.
What Yes Sounds Like, and Why Nothing Here Is Failure
Asked when a customer genuinely means they would pay, the answer is a description of a moment. It comes after many conversations and variants, when you return with something considered.
Mark: “The way to test if you’re right is the person doesn’t even need you to explain it.”
The example comes from Bloomfield Robotics, which used imaging to inspect specialty crops. A camera mounted on anything that moves runs the rows, images what it sees, and geolocates it, so a vineyard knows the condition of every vine continuously: grape count, average color, size, leaf density. A vintner knew the condition of every vine continuously, and the company reached farms on three continents before selling to Kubota. Early on, none of that was clear. After many rounds of back and forth, they took a solution to the CEO of a large global vineyard.
Mark: “I was one minute in and he said, stop. Here’s what you do. I’d never met this guy. And he laid out exactly what we did. He said, you probably do this, then this, then that, right? Am I right? And he said, how soon can we try all this?”
That is the signal. Now you are in business.
A prospect who finishes your pitch has recognized their own problem in your description, which enthusiastic feedback can never substitute for. Asked whether he followed this process in his own companies, the answer is candid.
Mark: “Almost all of them. You know why? Because it’s hard, it’s boring. Who wants to interview people? I want to build something. Who wants to get on a phone call with a farmer in upstate New York talking about vines? If you don’t want to do that and you want to get into ag, you’re gonna have a real problem, pal.”
On failure. Founders who discover their original theory was wrong tend to describe it as wasted money and lost time. The framing gets rejected outright.
Mark: “You didn’t make any mistakes. You just learned. There is no failure in entrepreneurship. It’s called learning. You’re doing something no one’s done before. If you’re beating yourself up every time the thing you thought was true was not true, you’re never gonna leave your home.”
On where AI opportunity sits. AI does lower the cost of product development, and the pitch for it deserves redirecting.
Mark: “I don’t believe the sale for AI is efficiency. I believe the real sale for AI right now going forward is effectiveness.”
A book on the subject is underway with Anand Rao, an AI professor at Carnegie Mellon, under the theme of AI that does real work. The distinction runs between the visible AI and the consequential kind.
Mark: “The AI we talk about is the fun stuff, humanoid robots and autonomous vehicles, or the scary stuff like Terminator. But a lot of the AI changing our world is deeply embedded in the infrastructure all around us that we never see.”
Efficiency gains come from automation, which predates this wave by decades. Effectiveness means machines doing things they have never done, extending the reach of human effort, and that is where the opportunity concentrates.
On the closing advice. A retired Navy SEAL turned entrepreneur supplied the framing.
Mark: “Entrepreneurship’s like being a Navy SEAL. You just gotta get comfortable being uncomfortable.”
The closing line follows from it: be hard on the problem and easy on yourself, because the journey is difficult enough already. Then go change the world.
Your Customer Discovery Checklist
Start with the first three.
- •Write down the problem you believe exists, then write the evidence that would prove it imaginary
- •Name the hundred people on earth who would feel that problem most acutely, and pick ten to call this month
- •Run every interview without mentioning your product, your technology or your company’s approach
- •Count how many of your last twenty customer conversations included a demo, and treat that as a warning
- •Apply the five whys to one interview until you reach something neither of you knew at the start
- •Sort your current work into project or business, and tell your team which it is
- •Ask what your product would need to be for a prospect to finish your pitch for you
- •Log every disproven assumption as learning with a date attached
Customer Discovery FAQ
What is the platinum hammer problem? Building an impressive technology and then searching for a problem it might solve. The term comes from Mark DeSantis and describes well-engineered products that satisfy no human need anyone will pay for, which is how technically excellent startups fail.
How many customer interviews are enough? Around a hundred, with the caveat that even a hundred may leave questions open. The first third typically reveals you are asking the wrong question of the wrong people. Depth arrives in the final quarter, once question and audience have both been corrected.
Why do customer interviews give misleading answers? Because people offer encouragement to founders they want to see succeed, and encouragement reads as demand. Presenting a solution during an interview compounds it, since the conversation then measures politeness. Discussing only the problem produces the honest answer.
What is the difference between a project and a business? Everything before product market fit is a project: a search for a problem worth solving. Everything after is a business, meaning you make something people buy for more than it costs to produce.
Ready to Find Out Whether Your Problem Is Real?
Most commerce and technology teams have a validation problem: a roadmap built on untested assumptions, customer conversations that generated warmth, and a product aimed at a need somebody assumed. CommerceShop works with brands and manufacturers on the systems behind growth, from customer research and conversion through to answer engine and generative engine visibility.
Book a Call With CommerceShop →Keep the Conversation Going
This is Episode 16 of Growth Files by CommerceShop, where operators and advisors share what works in commerce, technology and AI.
