Every time I talk to a founder for the first time, I ask them the same question. What’s your ideal customer profile?
I’ve asked it hundreds of times. I’ve almost never gotten a clean answer.
“All CMOs.” “Small companies and large enterprises.” “Anyone using ServiceNow.” “IT or security people at big regulated companies.” Every one of those sounds reasonable. None of them is actually an ICP. They’re a wish list. And building a business around a wish list is one of the most expensive mistakes I see founders make, including myself, twice.
At my first company, we had real traction with very small businesses, under ten employees. Then we landed one customer with 100 employees and convinced ourselves that was the path to faster growth. So we rebuilt the product for them. Changed how we sold. Hired more senior reps. And in the process pushed away the small customers who had gotten us there in the first place, while never fully building what the bigger customers actually needed because better-funded competitors were already ahead of us there.
At my last company, we were genuinely great at selling to IT teams. Then a few security leaders told us they loved our security features. So we decided to go after IT and security simultaneously. And when we tried to straddle that line, we lost the thing that made us special. Every piece of messaging had to leave room for two different buyers instead of one.
At the second company, an investor literally forced us into a formal ICP exercise before I finally saw what a narrow customer profile actually does. More deals close. Customers stay longer. The sales team stops chasing the wrong logos. The product team stops building features for customers you don’t even want. And suddenly the whole company is moving in one direction instead of three.
Once I went through it myself, I started seeing it everywhere. Every founder, every CEO, every company I’ve advised has had some version of this problem. Not because they’re careless. Because the incentives all pull in the wrong direction. Investors ask how big your TAM is. Buyers outside your target tell you they’d love your product. Both of those things pull you wider exactly when you need to go narrower.
A broad ICP doesn’t just slow you down. It drains you. It produces messaging that speaks to nobody. It builds products for customers you shouldn’t want. And by the time you realize what happened, you’ve already burned the runway proving it.
So I want to walk through what actually fixing this looks like, using the people who’ve taught me the most about it.
Start with a diagnostic, not a definition
Marcus Ryu, co-founder and former CEO of Guidewire, now invests and runs offsites with his portfolio founders. At one of those offsites, he runs a simple exercise. Write your ICP on a whiteboard. Say it out loud in front of your peers.
Almost nobody gets through it cleanly.
He’s watched this exercise expose more incoherence than almost anything else. Founders who are certain they have an ICP sit down to write it, and somewhere in the middle of externalizing it, they realize what’s actually on their list.
“The founders, when I ask them, tell me, ‘Yeah, we have an ICP, we have it.’ Then we start writing it on the whiteboard. And it turns out this is six ICPs, not one.”
That’s the real problem. It’s not that founders are lying. It’s that an ICP that only lives in your head, one you’ve never written down or said out loud to someone who can push back, isn’t actually one thing. It becomes whatever you talk yourself into on the sales call you happen to be on that day. The whiteboard exercise doesn’t fix the ICP. It just forces you to confront what you actually have.
Narrow until you can name every company on the list
Once you’ve written it down, the next question is how narrow is narrow enough. The answer I come back to more than any other came from Tom Buiocchi, the former CEO of ServiceChannel. He calls it “segment until it hurts,” and the best way to understand it is to hear exactly what he went through.
ServiceChannel’s target market was retail. That was it. Retail. And so their pipeline filled up with Walmart, two bakeries on the Upper West Side, a bagel shop on 79th Street, a small restaurant in Belgium, and McDonald’s, which is largely franchised. Their close rate was terrible. They were wasting enormous time on accounts that had no chance of closing.
So they started adding filters. Retail, but only food or apparel. Only 100 locations or more. Only multi-state, nationally distributed, because that’s where their value proposition actually kicked in. Only company-owned, not franchised, because franchises don’t have centralized decision-making. Only chains with centralized facilities management. Tom uses Costco as the example of a company that looks right but falls out on this filter. Every Costco has its own facilities manager operating independently. No one can make a decision for all of them. So Costco gets cut. And they wanted brands where consistency across every location actually mattered, which is how an Apple Store makes the list and Dollar General doesn’t.
By the time they finished adding filters, they had 682 companies in the entire country. And they knew every single one of them by name.
“Don’t fish in an ocean. Fish in a pond, and fish in a pond where you know the names of all the fish.”
That’s the real test behind Marcus’s whiteboard exercise. Not “do you have an ICP.” Can you literally name the companies in it?
Your ICP isn’t one group. It’s several smaller ones.
Craig Rosenberg has spent his whole career inside go-to-market motions, and now works with portfolio companies at Scale Venture Partners. He takes the “name the fish” idea one step further.
Even after you’ve defined your ICP tightly, the real unlock is breaking it into even smaller groups. A single ICP is made up of subsegments with different pain points and different buying triggers. If you try to write one message for all of them at once, even with AI helping you write it, the output is still generic because the input is too broad.
Craig’s tactical floor for an early-stage founder: have ten real conversations a week with people inside or near your assumed ICP. Not surveys. Not demos. Conversations where you’re listening, not pitching. You need enough repetitions to actually know whether you have the right ICP, what the sub-groups inside it look like, and what messaging lands with each of them. You can’t get that from a spreadsheet.
Your users and your customers are not the same thing
This is the part of the ICP conversation people skip most often, and it might be the most expensive mistake on this list.
Zach Smith ran Packet, an infrastructure company. Developer users loved the product. There were a lot of them. Revenue was coming in. His board thought they had figured it out.
Zach kept telling them they were wrong.
“Those are just my users who are paying to try and tell me about the thing. They’re not my customers. That’s not our target. They’re paying, I value them, I’m going to support them. But that is not where I’m going.”
Packet’s actual customer was the top 1,000 most technology-driven enterprises in the world. That’s who they were building for. The individual developers using the product were early signals, sometimes future champions inside those enterprises, but not the business itself. Zach had the discipline to tell his own board that the revenue in front of them wasn’t the revenue that mattered, which is one of the hardest things to say out loud when a number looks good on paper.
One of his first users was two people at a young company called HashiCorp. Not a customer yet. But the kind of company that could become one. He kept his eye on the top of the cake, even while building the layers underneath.
Champions are not customers
Melanie Fellay, CEO and Co-Founder of Spekit, learned a version of the same lesson from a slightly different angle.
Her team invested heavily in building a community of people who loved the product. They were enthusiastic, engaged, spreading the word. They were also almost entirely too junior to control a budget. Many of them never bought anything.
“A lot of those people never bought the products. Many of them were actually too low level in the organization to have the kind of budget. And I learned that the painful way.”
That’s the champions trap. Someone can love your product, amplify your content, and never sign a check, because the authority to sign isn’t sitting where the enthusiasm is. And Melanie’s second filter matters just as much. Her team was also selling to companies that were too small, 30 or 40 employees. They had the pain but not the structure or resources to do anything about it. The floor that actually worked was 150 employees, going up to around 500. Below that line, even willing buyers couldn’t make it stick.
Company size isn’t about who can use your product. It’s about who feels enough pain, and has enough organizational capacity, to actually fix it.
What it actually costs to narrow your ICP
Most founders talk about focus as if it’s a mindset choice. Sean Henry shows what it actually costs.
Sean co-founded Stord as a teenager. The company started as an Airbnb for warehouse space, open to anyone who wanted to store anything. Movie sets. Cars. Random stuff. People showed up, revenue came in, and none of it was the business they were trying to build.
So they shut it down. They fired the customers. A few hundred thousand dollars a year in revenue, walked away from at 18 years old.
“We spent the next six months so mad at ourselves, like, what did we just do?”
What came next was harder. They got into an accelerator with deep roots in freight brokerage, whose founders had built a business on cold calls and cold emails. The advice they gave Sean and his co-founder was simple: lock yourselves in a room and make hundreds of calls and thousands of emails a day. No budget for leads, so they built a scraper tool to pull contact information from government shipping records. Every container entering the US has a bill of lading. They went through hundreds of millions of documents and started outreaching at a scale most people wouldn’t attempt.
Their first real customer under the new model was a publicly traded solar panel manufacturer. Not the kind of customer an 18-year-old dreams about landing. Exactly the kind their actual business needed.
That’s what commitment to an ICP looks like in practice, not a slide, not a strategy session, real dollars, real time, real sacrifice before you have any cushion to absorb it.
Your ICP will need to change. That’s not failure.
Rohyt Belani, Former CEO and Co-Founder of PhishMe, makes the point that pulls everything together. His company started purely enterprise. Over time, they moved downstream into the mid-market. He didn’t treat that as a sign the original definition was wrong. He treated it as something you should expect to happen as the product and the market evolve.
What mattered was that they kept their eye on it. ICP wasn’t something they defined once and filed away. It was something they revisited intentionally, and every time it shifted, they made sure product, marketing, and engineering all moved together. That alignment is harder to maintain than the definition itself.
His closing point is the one I keep coming back to: if you don’t know who your ideal customer is, how are you going to decide anything else? Not your roadmap. Not your messaging. Not your next hire. Every downstream decision depends on that one upstream answer being clear.
Here’s where to start
Pull up your actual customer data, not your pipeline. Sort by retention rate and sales cycle length separately, and look at which segments sit at the top of both. Talk to your two or three best salespeople about the deals that felt like a fight versus the ones that just clicked. You’re looking for the small group where you win fast and they stick around. That’s your pond.
Then have the discipline to fish there, even when a logo outside it comes calling.
Guests Featured in This Episode:
Marcus Ryu, Co-Founder & Former CEO, Guidewire — LinkedIn
Tom Buiocchi, Former CEO, ServiceChannel — LinkedIn
Craig Rosenberg, Chief Platform Officer, Scale Venture Partners — LinkedIn
Zach Smith, Co-Founder & Former CEO, Packet — LinkedIn
Melanie Fellay, Co-Founder & CEO, Spekit — LinkedIn
Sean Henry, Co-Founder & CEO, Stord — LinkedIn
Rohyt Belani, Co-Founder & Former CEO, PhishMe — LinkedIn
Chapters:
0:00 — Trailer
0:55 — Why your ICP is probably too broad
5:55 — The exercise that exposes a fake ICP
8:59 — “Segment until it hurts”
14:00 — Why your ICP is actually several smaller ones
17:22 — Your users aren’t always your customers
20:08 — The real cost of firing the wrong customers
23:36 — Champions aren’t customers
24:59 — Your ICP will change, and that’s not failure
27:25 — Find your pockets of gold
