The ICP You Think You Have Was Written Before You Had Any Customers

Most ICPs are written in a conference room, before a single deal is closed, by people who are mostly guessing.

That's not a criticism. It's the only rational way to start. You build a hypothesis about who needs what you're building, describe that person in enough detail to give your sales team something to point at, and go find out if you're right.

The problem isn't that the founding ICP is a guess. The problem is what happens next — which, in most companies, is nothing. The ICP gets ratified by the first few wins, embedded into the pitch deck, and treated from that point forward as settled truth. It stops being a hypothesis and becomes a founding document. And founding documents are very difficult to challenge without it feeling like you're questioning the strategy itself.

So the ICP stays. The market moves. And somewhere in the widening gap between the two, CAC starts climbing — which almost nobody attributes to ICP drift. The diagnosis lands somewhere more comfortable: Marketing is spending inefficiently. The product is missing features. The sales cycle is too long. These are real problems, often. But they're also the symptoms that ICP misalignment reliably produces. Treating them without examining the ICP is expensive and temporarily satisfying, like turning up the heating because the window is open.

Here's the link that most GTM teams miss: your ICP is the input to your GTM motion, not a downstream output of it. When the ICP shifts — when the customer who stays and grows starts looking different from the customer your motion was designed to find — the GTM motion needs a version bump. Not a campaign refresh. Not a new persona card. A structural update to who you're pointing the whole machine at, and how you're reaching them. The ICP diagnostic is what triggers that versioning. Which means an unexamined ICP isn't just a targeting problem. It's a reason your GTM never gets past v1.0.

The Telco Vendor Problem

Nowhere is this pattern more visible than in enterprise technology vendors selling into telecom operators.

For decades, companies like Ericsson and Nokia built their go-to-market motion around a well-understood buyer: the CTO, the network architect, the head of radio access. These were the people who controlled infrastructure budgets, understood technical specs, and sat on vendor selection committees. The ICP was drawn around them because they were the ones who signed.

Then private 5G created an entirely different buying centre. The decision-maker for a private network deployment in a manufacturing plant or a port wasn't the network architect — it was a VP of Operations, a Head of Digital Transformation, sometimes a supply chain executive running a productivity case. At enterprises, IT and OT buyers are separate, with potentially divergent business priorities that sometimes compete for budget. TBR These buyers didn't appear in the original ICP. They didn't attend the same conferences. They didn't respond to the same messaging.

Nokia established deep partnerships with C&SI firms such as Accenture, Deloitte, EY, and Kyndryl for private network opportunities, and these relationships drove significant deal flow. TBR Those SI relationships gave Nokia a structural path to buyers the traditional telco GTM had never been designed to find. Ericsson, with a comparably strong product, made little mention of C&SIs in its messaging, with much more emphasis on CSPs and traditional resellers TBR — and found itself pointing the right product at the wrong buyer.

When growth stalled, the diagnosis pointed at product gaps and pricing. The ICP was rarely the suspect. It almost never is.

What actually needed to happen was a GTM version bump — a deliberate update to who the motion was designed to reach, which channels it relied on, and which buyer profile it was built to convert. Nokia got there through its SI channel strategy. Ericsson is still navigating the transition. The product delta between the two companies is marginal. The ICP and channel gap is not.

What Retention Data Reveals

The ICP built on closed-won optimises for the customer you can find and persuade. The ICP built on retention and expansion data optimises for the customer you can keep and grow. These are different customers more often than most GTM teams are willing to admit.

Net revenue retention is an unusually honest signal precisely because it strips out the deals that looked good on paper but represented the wrong fit. The accounts with the highest NRR — the ones that expanded without being resold to, that sent referrals, that brought your product into new parts of their organisation — share characteristics that frequently weren't present in the original ICP definition, because they weren't visible in closed-won data.

Sometimes it's a specific sponsor profile: not just a job title, but a combination of organisational mandate and internal credibility that determines whether your product gets adopted or shelved after the contract is signed. Sometimes it's an operational maturity signal — the accounts that expanded weren't the biggest ones; they were the ones with the internal discipline to extract value from the product rather than waiting for the value to extract itself. Sometimes it's structural: the accounts that stayed were the ones where your product solved a problem that had a budget owner, not just a champion.

None of these signals are visible in closed-won data. They live in renewal conversations, expansion reviews, and the notes your CS team writes after a difficult QBR that nobody in marketing ever reads.

The ICP Diagnostic: A Working Template

What most PMM teams actually need isn't a process — it's a sharper tool. Something that makes ICP drift visible before it becomes a revenue problem, and specific enough to trigger a genuine GTM version update rather than a deck refresh.

The diagnostic has four layers:

Layer 1 — The Founding Hypothesis What the ICP was designed around at inception. Firmographics, assumed buyer role, assumed pain point, assumed trigger event. Most companies have this written down. Very few have explicitly marked it as a hypothesis rather than a fact. Stress test: When was this last compared against actual retention data — not win rates, but expansion rates?

Layer 2 — The Closed-Won Profile Who is actually buying. Pull the last 12–18 months of won deals and describe them honestly: the actual job titles involved, the actual trigger that initiated the process, the actual objections that almost killed the deal. Compare this to Layer 1. The gaps are instructive — but they're still only telling you who you can sell to, not who you should be selling to. Stress test: Where does closed-won diverge from the founding hypothesis — and is that divergence signal or noise?

Layer 3 — The Retention Profile Who is actually staying and growing. Segment your base by NRR and look at the top 20%. What do those accounts have in common that isn't captured in Layers 1 or 2? This is the most important layer and the most consistently skipped, because it requires CS data most PMM teams don't have easy access to — which is itself worth fixing. Stress test: What characteristics predict expansion that your current ICP definition doesn't describe? That gap is your next GTM version.

Layer 4 — The Churn Profile Who left, and why — not the stated reason in the exit survey, which is almost always sanitised. The real reason: the account whose internal champion left and the product had no second owner; the account that renewed once out of inertia and never actually implemented; the account that bought based on a capability that turned out to be peripheral to what they actually needed. Churn is the ICP's immune system. It tells you who was always the wrong fit, even when the deal looked right. Stress test: What do churned accounts have in common that your ICP should have flagged as a risk signal at the point of sale?

The metric that ties all four layers together is NRR delta by ICP fit score. Score your accounts against your current ICP definition and compare NRR across high-fit and low-fit cohorts. If high ICP-fit accounts don't meaningfully outperform low ICP-fit accounts on NRR, the ICP isn't a fit model — it's a target list dressed up as a framework.

And when that gap does emerge — when Layer 3 describes a customer your current GTM motion wasn't built to find — that's not just an ICP update. That's a versioning trigger. The customer profile changed. The channels that reach them are different. The message that converts them is different. The motion needs to reflect that, deliberately and structurally, or the ICP update sits in a document while the GTM keeps running the old play.

The Organisational Trap

There's a reason this diagnostic rarely gets run. The data it produces is uncomfortable for multiple functions simultaneously.

If Layer 3 reveals that the best-retained accounts look different from the accounts the current ICP targets, that's a problem for Sales (chasing the wrong profile), Marketing (building campaigns around it), and Product (prioritising a roadmap shaped by the wrong customer's feedback). Nobody owns the ICP update because everybody would share the cost of acknowledging it.

The PMM who runs this diagnostic and surfaces the findings clearly is doing some of the most strategically valuable work in the organisation. They're also doing work nobody explicitly asked for, that challenges assumptions held by people more senior than them, and that produces no immediate pipeline impact.

Which is, perhaps, why most ICPs remain unexamined founding documents — a portrait of a customer someone described in a conference room, before the first deal was ever closed, that quietly shapes every GTM decision made in the years that follow.

The question isn't whether your ICP needs updating. It's whether the gap between your Layer 3 and your Layer 1 is wide enough to justify a GTM version bump — and whether anyone in your organisation has both the data and the standing to make that call.

Next: Blog 3 — PMM is the last human moat. And everyone is fighting over who owns it.

And if you haven't read Blog 1 yet — your GTM motion has a version number. The ICP diagnostic is one of the primary things that changes it.