The Last 50 Customers ICP Loop
Stop guessing who your customer is — your last 50 deals already told you.
- Difficulty
- Easy
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 90%
Clary's answer to 'how do I find my ICP?' is deliberately unglamorous: look at your last 50 customers. On day one you're making educated guesses from market and competitive research, but once you have history, the ICP is no longer guesswork — you identify the common threads across the companies you actually sold to. Company-level attributes (size, industry, geography, public vs private, revenue threshold, headcount) give you the ICP; the person who signed the PO in each of those deals gives you the buyer persona. He points to Snowflake, which built a constant feedback loop into its sales system, re-evaluating its ICP against its last 50 enterprise customers on a rolling basis. If that's too heavy, run it quarterly.
Origin
Clary formalised this from enterprise sales practice where ICP and buyer persona were operational terms, and cites Snowflake — the unicorn success story — as the proof case: they built a process where every 50 customers they'd reevaluate their ICP, making it a constant feedback loop inside the sales system rather than a one-off exercise.
Core principles
- 01Day one your ICP is an educated guess; after 50 customers it's data.
- 02The ICP is company-level; the buyer persona is the individual who signs.
- 03Common threads across closed customers beat market theory and competitor research.
- 04The loop must be recurring — ICP drifts as the product and market move.
How to run it
- 1
Start with educated guesses on day one
With no customers, look at the market and do competitive research — see who your competitors are targeting. Accept that this is a hypothesis, not an ICP.
Watch out Don't mistake the day-one guess for a finished ICP and defend it for years.
- 2
Pull the last 50 customers
Once you have history, list your last 50 closed customers. This is the evidence base that replaces guesswork.
- 3
Extract the company-level common threads
Tag each customer on any differentiator that identifies the company: size, industry, geography, public or private, government or private, revenue threshold, employee headcount. The shared attributes are your ICP.
Pro tip Notice what you're NOT winning — 'just sell to the biggest companies' ignores that public giants mean longer procurement and government means RFP.
- 4
Identify who actually signed
Within each of those companies, find the person who made the decision — their characteristics and job title. CMO, Director of Marketing, COO, CEO, VP Finance, CTO. That's your buyer persona.
Watch out The person you talked to most isn't always the person who signed.
- 5
Build shareable avatars
Create avatars with faces in a deck or database representing these people — education, salary, likely location, observed traits. The point is org-wide alignment on who you're selling to.
Pro tip You won't figure out everything; you only need enough to form an educated opinion that a new contact looks like your last 40-60 buyers.
- 6
Make it a loop, not an event
Re-evaluate on a rolling basis — Snowflake did it every 50 customers as a built-in feedback loop. At minimum, do it once a quarter against your past 50.
Watch out A stale ICP quietly misdirects every campaign downstream of it.
In the wild
Snowflake built a process into their sales system where every 50 customers they would reevaluate their ICP, treating it as a constant feedback loop against their last 50 enterprise customers rather than a fixed definition set at founding.
→ Clary cites them as a huge success story and unicorn, using the loop as the reason their targeting stayed accurate as they scaled.
Clary walks through the instinct to just target the largest accounts: if you sell to a public company with over $500 million in revenue, there's a longer procurement process; if you sell to government, you have to go to RFP.
→ The 50-customer data tells you which of those trade-offs your product actually survives, replacing the instinct with evidence.
Common mistakes
Treating ICP as a one-time exercise
Set once at founding and never revisited, the ICP drifts away from who's actually buying and silently misdirects the whole go-to-market.
Defaulting to the biggest logos
Bigger companies carry longer procurement and RFP overhead. Size is not fit, and your closed-won data will show it.
Confusing the ICP with the buyer persona
The ICP is the company; the buyer persona is the individual who signs. Collapsing them leaves you targeting a company with no idea who to email.
Is it for you?
Best for
Companies with at least a few dozen closed customers who want to replace ICP guesswork with observed data.
Not ideal for
Pre-revenue startups with no customer history — they're still in the educated-guess phase Clary describes.
From the transcript
“You literally can look at your last 50 customers. And from those last 50 customers, it's no longer guesswork. You can now find what your…”
“Funny enough, the company's Snowflake. Huge success story, unicorn. They built this process where every 50 customers, they would reevaluate their ICP. So it was…”
“Then within that company, then you look who is the person who made the decision in that company? Who's the decision maker? That's the buyer…”
From the episode
Scott D. Clary: Sales Strategies That Close Deals
Scott D. Clary