The short version:
- Most ICP documents describe a market segment. They tell you who could theoretically buy, not who to email on Monday morning.
- A tight ICP produces consistent yes/no answers in under 60 seconds per account. If two people on your team apply your ICP independently and build different lists, the definition is not doing its job.
- The best ICPs include a triggering condition (a recent event or behavioral signal) and a meaningful disqualifier, not just firmographic ranges.
- Build backward from your 10 best closed accounts. The shared attributes you find are almost always more specific than what is in your current ICP document.
A founder at a 15-person fintech startup showed me his ICP document last spring. Two slides. "Mid-market B2B companies, 50 to 500 employees, Series A through C, sales-led growth, need better pipeline visibility."
There were roughly 40,000 companies in the US that fit that description. He had budget for 300 emails per month.
His ICP was not a profile. It was a market size.
The Difference Between a Market Segment and an ICP
A market segment says who could buy. An ICP says who should hear from you this week.
"VP of Sales at a SaaS company with 50 to 300 employees" is a segment. It has tens of thousands of members.
"VP of Sales at a SaaS company with 50 to 300 employees who inherited a BDR team from a predecessor, missed quota two quarters in a row, and has a renewal or hiring decision coming in the next 60 days" is closer to an ICP. It has a few hundred members at any given time.
The second definition tells you something about urgency. The first does not.
Most sales teams have the first kind and call it the second kind.
The Two-Person Test
Here is a fast way to check if your ICP is actually working.
Take your ICP definition and a list of 20 company names. Give both to two people on your team independently. Have each person mark every company as "yes, reach out" or "no, skip." Compare results.
If they agree on 17 or more, your ICP is functional. If they disagree on five or more, it is not a filter. It is a description that each person is interpreting differently.
A useful ICP produces consistent decisions across people who have never talked to each other about it. If it does not do that, every rep is building a slightly different list, and none of the lists are right.
What Tight ICPs Have in Common
The strongest ICP definitions share three attributes.
A triggering condition, not just a profile. Not "B2B SaaS with 50 to 200 employees" but "B2B SaaS with 50 to 200 employees that crossed three sales hires in the last six months or just expanded into a second geographic market." The condition narrows the pool from thousands to hundreds and gives you a reason why now is different from six months ago.
A meaningful disqualifier. A tight ICP does not just say who fits. It says who almost fits but should be skipped. "Requires an enterprise procurement process" is a disqualifier if your sales cycle tops out at 30 days. "Series D and above" is a disqualifier if your product is priced for growth-stage budgets. The disqualifier is what keeps you from burning accounts that look right but will not close.
Roots in actual closed deals. Not in who you hope to sell to, but in who bought. A professional services firm doing $8 million in revenue had an ICP that read "marketing agencies, 10 to 50 employees, US-based." When they pulled their 12 best customers and looked for shared attributes, 11 of them had hired a new operations director within 12 months of signing. That one attribute, which never appeared in their ICP document, was the best predictor they had. They added it. Reply rate on outbound went from 6% to 14% in 60 days.
How to Build Backward From Your Best Accounts
Pull your 10 best closed-won accounts. Define "best" however makes sense for your business: highest ARR, lowest churn, fastest time to close, most referrals given.
For each account, write down everything that was true about them at the point they signed. Not now, but then. Employee count, revenue range, the role that made the decision, what tool they were replacing, what had changed in the business in the 12 months before they talked to you.
Then look for overlap. Which attributes appear in eight or nine of your 10 best customers but not in your five worst deals or your five deals that stalled and went nowhere?
That overlap is your real ICP. It is almost always more specific than the document you wrote from first principles.
A DevOps tooling company did this exercise and found that every one of their best accounts had gone through an ISO 27001 audit in the previous 18 months. That is not a demographic. It is a behavioral signal that reliably predicted purchase. They added it to their ICP and started monitoring for companies completing or approaching that audit.
The 60-Second Test
Here is the practical calibration. Pull up a company's LinkedIn page, their website homepage, and their current job postings. Set a 60-second timer. Make a yes/no call.
If you cannot make that call in 60 seconds, your ICP requires information that is not readily available in public sources. Which means your team is spending 20 minutes per account before knowing whether it is worth reaching out. That research time is your prospecting budget, and most of it is being spent on accounts that should have been filtered out at the first pass.
A good ICP uses criteria you can confirm quickly: headcount range, industry, whether a specific team or function exists, recent hires, recent announcements. If your criteria require deep internal research before you can evaluate a company, the criteria are wrong ones for an ICP.
What a Wide ICP Actually Costs You
A vague ICP does not just waste emails. It wastes accounts.
Reaching out to a company that is not ready, or not a fit, does not just produce a non-reply. It produces a record. The contact now has your name in the wrong context. If that company becomes a real fit 12 months later, the first impression is still there.
A growth team at a $15 million ARR SaaS company tracked this directly. Accounts that received cold outreach before they hit the company's qualifying criteria were 40% less likely to convert later, even when they eventually became exact ICP fits. The premature outreach created a reputation signal that persisted.
Your ICP is not just a list filter. It is a protection mechanism for the accounts that will matter next year.
The Revision Worth Making This Week
Take your ICP document. Read every criterion. For each one, ask whether it narrows your market from thousands to hundreds, or whether it just describes the category.
If most of your criteria are category descriptors, add two things: one triggering condition (a recent event, a behavioral signal, something that makes now different from six months ago) and one meaningful disqualifier (something that removes the almost-right accounts).
Then run the two-person test. If your ICP produces consistent answers, it is working. If it does not, it is a market segment wearing a new name.
TL;DR:
- A market segment tells you who could buy. An ICP tells you who to email on Monday. Most teams have the first and call it the second.
- If two people apply your ICP independently and build different lists, the definition needs work.
- The best ICPs include a triggering condition and a meaningful disqualifier, not just firmographic ranges.
- Build backward from your 10 best closed accounts. The shared attributes you find are almost always more specific than your current ICP document.