There’s a simple way to know if your ideal customer profile is doing its job: ask it five questions.
If your ICP can answer them without flinching, it’s probably doing real work in the field. If it can’t, you have an ICP problem the org chart won’t reveal. Reps are routing around the document, deals are slipping in ways nobody can predict, and the next QBR is going to be uncomfortable.
I run this test in pipeline reviews. It surfaces more in fifteen minutes than most ICP refreshes surface in a quarter. Here’s the test, and how to build an ideal customer profile that passes it.
In this article:
- Five questions your ICP should answer
- How to build an ideal customer profile that passes the test
- Why most ICPs fail the test anyway
- Key takeaways
Five questions your ICP should answer
Pull the doc out of the shared drive. Read it. Now answer these.
1. Can you explain why your last 10 closed-won customers actually bought now?
Not “they fit the profile.” Not “the timing was right.” Specifically: what changed in their world that turned a long-time prospect into a signed contract this quarter?
If your answer is firmographic (“mid-market fintech, 200 to 500 employees, US-based”), your ICP is doing half the job. Firmographics describe the company you’d sell to. They don’t explain why anyone bought. The companies that signed had a trigger: new exec, funding round, regulatory deadline, competitive failure, board mandate. One deal I watched sat flat for a year and closed in 3 weeks the moment a new CISO walked in with a board mandate to consolidate vendors. Nothing about the firmographics changed. The trigger did. If your ICP doesn’t capture trigger patterns, your reps are flying blind on the question that matters most: why this account, why now?
2. Who owns the ICP, and when was it last updated?
Check the last-modified date.
If it’s been more than six months, the document is probably wrong in ways you haven’t measured. B2B contact data goes stale fast, and senior buyers turn over fast. Fortune 500 CMOs average 4.3 years in role, the shortest run of any C-suite seat and below the 4.9-year C-suite average (Spencer Stuart’s 2024 CMO Tenure Study). A two-year-old ICP is targeting buyers who, in many cases, aren’t in those seats anymore.
Owner matters more than cadence. If no name is on the document, no update is happening on purpose.
3. If you pulled your last 20 closed-lost deals, how many would technically match the ICP?
This is the cheapest research you have. If most of your lost deals technically fit the profile, the ICP isn’t filtering for the right things. Either the firmographic layer is too loose, or the behavioral layer doesn’t exist.
Most teams have never run this exercise. They build the ICP from closed-won and never check it against closed-lost. The document keeps validating itself while the win rate quietly drifts.
4. Are reps prioritizing accounts by the ICP, or by their gut?
Watch what your top reps do for one week. Are they working accounts in priority order from the ICP-driven list? Or are they working a different list, the one they’ve decided will close?
If reps are routing around the ICP, one of two things is true. Either the ICP is wrong and reps know it before strategy does. Or the ICP is right and reps don’t trust the data behind it. Both are problems. Both get worse the longer they go unaddressed. This is the same strategy-to-field disconnect that breaks most GTM plans, playing out at the account level.
5. Can you draw a line from a market shift in the last six months to a change in your ICP?
Markets move every quarter. A new competitor launches. A category leader pivots. A regulation passes. A vertical opens or closes. If none of those shifts have resulted in an update to your ICP, the document is operating in a market that no longer exists.
You don’t need to update the ICP every time something moves. You need a defined trigger for when an update happens. If you can’t name the last trigger that fired, there isn’t one.
How to build an ideal customer profile that passes the test
If your ICP failed two or more of those questions, it’s time for a rebuild. Here’s the sequence I’d walk a team through:
Start with your healthy customers, not your assumptions
Pull 18 to 24 months of closed-won data. Then narrow further: which of those customers are still active, expanding, and reference-ready? Build your ICP from that subset, not the full closed-won list. Your healthy customer base tells you the truth, your sales deck won’t.
PostHog made this point well when they redefined their own ICP: you know it’s correct when accounts inside it activate and retain materially better than the ones outside it. Most teams never check the second half.
Layer behavioral signals over firmographics
For each account in your prospect universe, you need both static fit (firmographic match) and dynamic readiness (in-market signals). Static fit tells you whether to consider an account at all. Behavioral signals tell you which of those accounts are worth your reps’ calendars this quarter.
New executive hires. Recent funding. Tech stack changes. Competitive losses. Public statements about strategic priorities. These signals separate the 200 accounts your reps should work on this quarter from the 2,000 that technically match the firmographic profile.
Pressure-test against closed-lost
Run the exercise from Question 3. Take your 20 most recent losses, score them against the ICP, and see what shakes loose. The firmographic criteria that show up in most of your losses don’t predict anything. They either need to be cut or paired with a behavioral filter that does the predictive work.
Wire it into how reps actually work
Tag every opportunity in CRM with an ICP-fit score. Look at the win rate by score in your weekly forecast calls. If reps are winning on accounts marked low-fit, the ICP needs to evolve. If they’re losing on high-fit accounts, the executional playbook does. Either way, you’ll know inside one quarter instead of waiting until the QBR.
Define an owner. Define a cadence. Define what triggers an off-cycle update. Without those three things, you’ve just done a one-time refresh and started the clock on the next slide-in-a-drawer.
(Image: Target Accounts. Liminal GTM interface showing filtered account list with scoring, firmographic filters, and signal-based prioritization.)
(Alt text: Liminal GTM Target Accounts dashboard showing ideal customer profile scoring with firmographic filters and behavioral signals)
Why most ICPs fail the test anyway
I’ve watched plenty of teams run a clean ICP rebuild and still end up with a slide in a drawer six months later. The reason is almost never the analytical work. The reason is organizational.
Most teams treat the ICP as a deliverable. Marketing builds it, presents it at SKO, files it in the shared drive, and moves on. By the time anyone references it again, the market has moved, and the document is wrong. Reps know. Strategy doesn’t.
A working ICP is an operating system, not a deliverable. It has an owner, a cadence, a feedback path from the field, and a place in your CRM where it actually shapes how reps prioritize their day. Without those four things, you don’t have an ICP. You have a slide.
Key Takeaways
- The way to know if your ICP holds up isn’t to read it. Test it against five questions about how it actually performs in the field.
- The five diagnostic questions: why did your last 10 closed-won customers buy now; who owns the ICP and when was it updated; how many of your closed-lost deals technically matched it; are reps prioritizing by it or by gut; what market shifts have changed it lately.
- If your ICP fails two or more, rebuild from your healthy customer base, layer behavioral signals over firmographics, pressure-test against closed-lost, and wire it into CRM.
- The biggest mistake is treating the ICP as a deliverable rather than an operating system with an owner, a cadence, and a feedback loop from the field.
Liminal turns your ICP into an operating system instead of a slide. Target Accounts surfaces the companies that fit right now, and the list refreshes as the market moves, so reps always know who to work. → Find out more
