Most ICP documents are fiction. They describe the customer the founder wishes they had, not the customers who actually closed and stayed and grew.
I’ve reviewed a lot of these documents. They tend to say something like: “Series A–B SaaS companies with 50–200 employees, facing a challenge around X, where the decision-maker is a VP of Y.” It sounds right. It might even be right in parts. But it was usually written in a strategy session, not derived from the data, and it’s been quietly obsolete for two years without anyone noticing.
Here’s how to do it properly.
Why most ICPs are aspirational rather than analytical
The ICP exercise gets done once, usually early in the company’s life, often when there isn’t enough data to do it well. The founder writes down who they want to sell to — based on instinct, based on the reference customers they’re proudest of, based on who they think the product is “really for.”
That document then gets inherited by the first marketing hire, cited in the sales deck, and gradually treated as fact.
The problem is that what you want to be true about your best customers and what is actually true are almost always different. The customers you’re proudest of are often not the most profitable. The vertical you’re excited about may have a longer sales cycle and lower win rate than the boring vertical you’ve been winning quietly for years. The companies you target because they have budget may churn faster than the ones that were harder to close.
Until you look at the data, you don’t know which of these is true for your business. And most founders resist looking because they don’t want to be told they should stop chasing the deals they find interesting.
The exercise: building an ICP from closed-won deals
Start with your closed-won deals from the last 18–24 months. You want a minimum of 20–30 deals to work with; if you have fewer than that, include lost deals for comparison purposes too.
For each closed-won deal, record:
- Company size (headcount and revenue band if you have it)
- Industry / vertical
- Geography
- Deal size (ACV)
- Sales cycle length (days from first contact to signature)
- Inbound or outbound (how they found you, or how you found them)
- Who the champion was (title, seniority, department)
- Who signed (economic buyer — same person, or different?)
- Time to first value (when did they actually start using the product meaningfully?)
- Current status (still a customer, churned, expanded?)
Now you can analyse. Group the deals and look for patterns across four dimensions:
- Win rate by segment — which company types did you close most often relative to how many you engaged?
- Average deal size by segment — where do the bigger deals come from?
- Sales cycle length by segment — where do deals move fastest?
- Retention and expansion by segment — which customers stayed and grew?
These four dimensions will often point in different directions. The segment with the best win rate may have the smallest deals. The fastest sales cycles may be in a vertical that churns. You need to weight these against each other based on what your business actually needs right now — more revenue quickly, or more sustainable ARR?
What attributes actually predict good customers
Most ICP frameworks focus on firmographics — company size, industry, geography. These matter, but they’re descriptive rather than predictive. The attributes that tend to actually predict deal quality are different.
Pain acuity is the most important one. Does this type of company have the problem you solve as an urgent, business-critical issue — or is it a nice-to-have for them? A £40k ACV deal where the customer genuinely cannot run their business without you is a better deal than a £60k ACV deal where you’re a line item that gets reviewed every quarter.
Budget ownership matters more than you’d think. In some company types, the budget for what you do sits clearly with the economic buyer you’re talking to. In others, it requires a procurement process, a committee, a board sign-off. The latter isn’t necessarily a bad deal — but it changes everything about how you sell.
Implementation complexity affects both sales cycle and churn. If getting to value requires a long integration project, that affects which companies are actually good fits for you at this stage. A complex implementation that goes badly is expensive for you and for the customer.
Prior category experience — has the buyer done this before, or are they buying your category for the first time? First-time buyers need more education, take longer to close, and sometimes churn when the reality doesn’t match their expectations. Buyers who’ve used a competitor have shorter sales cycles and clearer success criteria.
None of these attributes are captured in a typical firmographic ICP. But they show up clearly when you interview your best customers and ask them directly.
The difference between ICP and persona
These get conflated constantly. They’re different things.
Your ICP is a company profile. It describes the type of organisation that is an excellent fit for your product or service — its size, structure, problem profile, and buying characteristics. It tells you which doors to knock on.
Your personas are the people inside that company — the champion who’ll advocate for you, the economic buyer who’ll sign, the technical evaluator who’ll raise objections, the end user who’ll determine whether they renew. Personas tell you how to navigate the conversation once you’re inside.
You need both. But you need the ICP first, because there’s no point developing detailed personas for a company type that doesn’t fit your profile.
Why founders resist this exercise
The ICP analysis will almost certainly tell you to stop pursuing some deals you want to pursue. It might show that a particular vertical you find exciting has a 12% win rate and churns at twice the rate of your boring, bread-and-butter segment. It might show that enterprise deals — which feel like validation — are actually destroying your sales efficiency compared to the mid-market.
Saying no to revenue is hard, especially at £500k–£2M ARR when every deal feels important. But an ICP that doesn’t help you say no isn’t an ICP — it’s a wish list.
The discipline is in treating the analysis as a decision-making tool, not a brand exercise. If the data says your best customers are 30-person professional services firms in the UK, that’s your ICP, regardless of whether you’d rather be selling to Series B scale-ups.
The ICP as a live document
Your ICP should be revisited every six months, minimum. As you add customers, the data shifts. As your product develops, the fit profile changes. As your go-to-market matures, you might deliberately move upmarket or into a new vertical — and the ICP needs to reflect that intentional shift, not trail behind it by two years.
Build it into your quarterly business review. Pull the closed-won data, run the same analysis, check whether the profile has changed. If it has, understand why before you update the document.
A well-maintained ICP is one of the highest-leverage tools in a B2B revenue function. It informs your outbound targeting, your inbound qualification, your marketing positioning, your sales process design, and your customer success priorities. Everything downstream gets cleaner when the ICP is honest.
If your qualification process is shaky — which it often is when the ICP is aspirational rather than analytical — my post on how to qualify beyond BANT covers the framework I use in practice. And if you’re trying to understand whether the sales process itself is the constraint, how to build a sales process walks through the full structure. If you’d like to work through the ICP exercise properly with someone who’s done it across a lot of B2B businesses, the Discovery Week is a good starting point.