The ICP Model That Survives Contact With Real Sales Calls

‍TL;DR: An ICP model built entirely from desk research, firmographic filters, market sizing, competitor analysis, looks rigorous and often doesn't match what happens on real sales calls. The fix isn't abandoning desk research; it's validating every assumption against a specific, recurring test: pulling the transcript or notes from the last ten sales calls and checking whether the stated ICP criteria actually predicted who bought, who stalled, and who never had a chance. An ICP that fails this test isn't wrong forever; it's wrong until it's been corrected against what real calls are actually showing.

‍

An ICP document built in a strategy workshop, complete with firmographic filters and a tidy slide, often survives exactly one encounter with reality: the first time a sales rep says, "that's not who's actually buying." Most ICP models are built once, presented confidently, and never checked again against what real sales conversations are revealing week to week.

Why desk-research ICPs fail on contact with real calls

An ICP built from market sizing data, competitor analysis, and reasonable assumptions about who should want the product is a hypothesis, not a fact. It's often a well-informed hypothesis, but it hasn't yet been tested against the messiest, most reliable source of truth available: what actually happens when a real rep talks to a real prospect. Desk research can tell you who plausibly has the problem your product solves. Only real sales calls can tell you who actually recognizes they have that problem, has budget to fix it, and moves at a pace that makes them a viable target now rather than in theory.

The specific test that catches a broken ICP early

Why closed-won deals are the most honest evidence available

Every closed-won deal is proof that a specific company, at a specific stage, with a specific problem, actually paid money to solve it. That's a stronger signal than any amount of market sizing or persona research, because it's evidence of real behavior rather than a prediction of it. Pulling the last ten closed-won deals and checking them directly against the documented ICP's stated criteria, company size, industry, tech stack, deal size, sales cycle length, is the single fastest way to find out whether the ICP is describing reality or describing an assumption that hasn't been tested yet.

Why stalled deals are just as informative as closed ones

A closed-won deal tells you who buys. A stalled deal, one that entered the pipeline looking like a strong ICP match and then went quiet, tells you what the ICP is missing. If several stalled deals share a specific, recurring objection, budget authority sitting with a different stakeholder than assumed, a competing internal tool the ICP didn't account for, a timing mismatch tied to a fiscal year nobody flagged, that pattern is a real disqualifying signal the ICP document should incorporate, not a coincidence to shrug off deal by deal.

Why rep feedback catches what CRM data alone misses

CRM data shows what happened; it doesn't always show why a rep chose to spend time on a particular lead or deprioritize another. A rep who's worked the market for months often has an accurate, if informal, sense of which "ICP-fit" accounts never actually convert despite matching every documented criterion, because of some factor the ICP document doesn't capture, an internal champion who lacks real influence, a procurement process too slow for the deal size to justify. This kind of pattern recognition rarely gets formally logged anywhere, which is exactly why asking reps directly, in a structured way, surfaces information a purely data-driven analysis misses.

Building the correction loop, not just the one-time check

A single check against these three tests is useful once, but the real value comes from making it a recurring habit rather than a one-time validation exercise. purple path's six-step framework for identifying your ideal customer profile covers the initial build process; the test in this article is the ongoing correction loop that keeps that initial build honest as real sales data accumulates. An ICP built once and never re-checked against actual outcomes drifts further from reality every quarter that passes without the correction.

Why this test matters more for a Series A company specifically

A Series A company at €10-30M ARR, still refining a sales-led motion, has less historical data to work with than a larger, more established company, which makes each individual signal, each closed deal, each stalled one, proportionally more valuable. Ignoring what ten real sales calls are revealing, in favor of trusting an ICP document built from external market research, wastes exactly the kind of direct, high-signal evidence a smaller company most needs to be paying attention to.

What to do when the test reveals a mismatch

Finding a mismatch between the documented ICP and actual sales outcomes isn't a failure; it's the process working as intended. The fix is a direct rewrite of the specific criterion that failed the test, not a wholesale abandonment of the ICP concept. If closed-won deals consistently skew toward a company size bracket smaller than what's documented, narrow the stated range. If a specific tech stack criterion never actually predicted anything about who bought, drop it and replace it with whatever pattern the real data does support. purple path's approach to mapping the buying committee first in ABM depends on this same kind of accuracy; an ABM program targeting the wrong buying committee because the underlying ICP was never corrected against real outcomes wastes the same effort a demand generation program would waste chasing the wrong accounts.

Why this test works better as a habit than as a one-time project

Treating this as a one-time validation exercise, run once when the ICP document is first built and never revisited, misses most of the value. Sales conversations keep happening every week, which means new evidence about what the ICP should actually say keeps accumulating every week too. A quarterly rhythm, pulling the most recent batch of closed and stalled deals and checking them against the current documented ICP, catches drift while it's still small and easy to correct, rather than discovering eighteen months later that the entire sales team has quietly stopped trusting a document nobody bothered to update.

What happens when sales stops trusting the ICP document entirely

A specific, damaging outcome worth naming: once reps notice repeatedly that the documented ICP doesn't match who's actually buying, they stop consulting it altogether and start relying entirely on individual gut feeling about who's worth pursuing. This isn't necessarily catastrophic in the short term, since experienced reps often develop genuinely good instincts, but it means the company loses the benefit of a shared, documented understanding that new hires can learn from, that marketing can target against, and that leadership can use to make consistent resourcing decisions. An ICP document that's actively maintained against real outcomes stays useful precisely because it earns the team's continued trust; one that's built once and left stale quietly loses that trust, and rebuilding it later takes considerably more effort than maintaining it would have.

Frequently Asked Questions

How many sales calls or deals are enough to run this test meaningfully?

Ten closed-won and a handful of stalled deals is a reasonable starting sample for an initial check. A company with a longer sales cycle and lower deal volume may need to look back further in time to gather enough data points for the pattern to be reliable rather than coincidental.

How often should this test be rerun once the ICP has been corrected?

Quarterly is a reasonable cadence for most Series A companies, since sales cycles and market conditions can shift meaningfully within that window, and a correction made once doesn't guarantee the ICP stays accurate indefinitely.

What if sales and marketing disagree about what the real ICP data is showing?

This disagreement is worth surfacing directly rather than resolving quietly in favor of whichever team has more say. Reviewing the actual deal data together, rather than relying on each side's separate impression of it, tends to resolve the disagreement faster than debating opinions without the underlying evidence in front of both parties.

Does this test work the same way for a product-led growth motion?

The underlying principle, checking assumptions against real behavior rather than desk research alone, applies broadly, but the specific data points shift: product usage signals and self-serve conversion patterns matter more than sales call notes in a PLG context, since there may be fewer or no sales calls to review directly.

Can an ICP built entirely from real sales data, with no desk research at all, work well?

It can work, particularly for a company with an established, mature sales motion generating plenty of real data to learn from. Desk research still adds value early on or when expanding into a genuinely new market segment where real sales data doesn't yet exist to validate against.

‍

Running this test against your own last ten sales calls is a half-day exercise that can save months of chasing the wrong accounts. Talk to purple path about validating your ICP against what your real pipeline is actually showing.