How to define your ideal customer profile (ICP): a practical guide

An ideal customer profile (ICP) is the short, data-backed description of the companies you win fastest, keep longest and serve with the least friction. It does not list who could buy; it tells you who to actively pursue so your sales effort becomes measurably more efficient.

Updated: June 2026

The gap is real. According to SiriusDecisions (now part of Forrester), companies with a clearly defined ICP achieve roughly 68% higher account win rates than those without one. An ICP is not a marketing ornament; it is a lever on revenue and speed.

This guide walks you through it step by step: how to derive your ICP from real data, turn it into testable criteria, and use it every day – with a criteria table, a worked example in CHF and a checklist you can act on.

What is an ICP – and what is it not?

An ICP describes the type of company (in B2B) or the customer segment that fits your offer best: industry, size, region, maturity. It is a filter rule for your market – not a description of one individual person.

The distinction from a buyer persona matters. The ICP says which companies you target; the persona says which people inside those companies you convince (role, goals, objections). They complement each other, but they are not the same thing.

• ICP: «Swiss accounting and consulting firms with 5–30 staff, project-based, no IT department.»
• Persona: «Managing director, wants less spreadsheet work, fears long rollouts.»

An ICP is also not an advertising audience («SMEs in Switzerland»). It is narrower, sharper, and it actively decides who you do not pursue.

Why is the effort worth it at all?

A sharp ICP saves time and money at both ends: marketing targets more precisely, sales stops wasting weeks on hopeless deals, and onboarding and support become more predictable because customers resemble each other. Less waste, more repeatability.

The Swiss context makes this especially relevant. According to the federal SME portal (kmu.admin.ch, 2025), there are around 600,000 active SMEs in Switzerland, and roughly half of all businesses use a CRM. If you sell into this market, you cannot address everyone – an ICP forces focus.

There is more: once you know your best customers, you can reproduce them. That is the core of an ICP – turning lucky wins into a repeatable pattern. For how a CRM makes that data visible in the first place, see What is a CRM.

Step 1: What data do you need as a foundation?

Before you define criteria, you gather facts from your own book of business. An ICP is derived from real winners, not from wishful thinking. Pull a list of your best customers from the last two to three years and add a few hard attributes per company.

Concretely, this list helps:

• Revenue / contribution margin per customer
• Sales cycle from first contact to close
• Retention (how long does the customer stay?)
• Support load (many tickets or few?)
• Industry, headcount, region, software stack

If this data is scattered across spreadsheets, inbox and calendar, that scatter is precisely the problem. A cleanly maintained CRM turns this step into minutes rather than days.

Step 2: Who are your best customers really?

In the dataset, identify the top 10–20% – not by gut feel, but by profitability and effort. The best customer is rarely the loudest; often it is the one who signs quickly, stays long and needs little support. Deliberately sort by these metrics.

A useful trick: form three groups.

1. Dream customers – profitable, loyal, low-maintenance. This is where you hunt for patterns.
2. Ordinary customers – fine, but not defining.
3. Problem customers – high effort, thin margin, churn early.

The problem group is gold: what do these companies have in common? Those very traits belong in your exclusion criteria later. An ICP always defines both – who you seek and who you avoid.

Step 3: Which criteria belong in the ICP?

Now you translate the patterns from your dream customers into concrete, testable criteria. Good ICP criteria are observable from the outside (industry, size) or easy to ask about (software, process maturity). Avoid vague adjectives like «ambitious» – you cannot filter on those.

Split the criteria into two layers: firmographic (what the company is) and behavioural (how it acts). A compact template:

Keep the list short: five to seven criteria are enough. An ICP that covers everything filters nothing.

Step 4: How do you weight and score the criteria?

Not every criterion matters equally. Assign points and a weight per criterion so the profile becomes a simple score. That lets you triage incoming enquiries fast, instead of re-debating every company. The goal is a single number everyone on the team understands.

A simple model: 0–2 points per criterion, multiplied by a weight of 1–3. Sum the points and define thresholds.

• High score (A): pursue actively, priority.
• Medium (B): handle when capacity allows.
• Low (C): politely decline or refer on.

Crucially, the model must stay simple, or nobody will maintain it. A score on the back of a napkin that actually gets used beats a perfect model that sits in a drawer.

Worked example: what does a wrong customer cost?

An ICP becomes tangible when you price the cost of ignoring it. Take an accounting firm with two salespeople that receives 20 enquiries a month. Without a filter, it chases all of them – including those that will never fit the profile.

A conservative model calculation:

Eight hopeless leads fewer per month save around CHF 2,160.00 of internal time – time that flows into the eight A-customers who actually close. The hourly rate here is an assumption; replace it with your real figures.

For how internal costs like these compare with CRM costs, we do the honest maths in CRM pricing models explained.

Step 5: How do you test and refine your ICP?

An ICP is a hypothesis, not a law of nature. Apply it consistently for four to eight weeks, then compare: do A-customers really close more often? Do they stay longer? If not, adjust the criteria. An ICP that is never reviewed ages fast.

Concrete checkpoints for the review:

• Win rate A vs. B vs. C – is the gap what you expected?
• Sales cycle – do A-customers close faster?
• Mis-scored cases – were there A-customers that flopped?

AI use among Swiss SMEs rose from 22% (2024) to 34% (2025), according to kmu.admin.ch (2025). That is relevant because modern CRMs increasingly help surface patterns in won and lost deals automatically – which makes ICP maintenance easier. For why clean inputs matter here, see Clean CRM data.

Step 6: How do you embed the ICP in daily work?

A profile sitting in a PDF folder changes nothing. To have effect, it must be visible where decisions are made: in the lead form, in the CRM field, in the weekly sales meeting. Make the ICP score a mandatory entry on every new contact.

Proven ways to embed it:

• ICP score as a field on every lead in the CRM.
• One short, shared definition (a single page, not twenty).
• A regular look at rejected C-enquiries – was the rejection right?

During a CRM rollout in particular, it pays to design the ICP in from the start – and to avoid the usual traps, which we cover in 7 CRM rollout mistakes.

Frequently asked questions

What is the difference between an ICP and a buyer persona?
The ICP describes the company (industry, size, region, maturity) that fits best. The persona describes the person in that company (role, goals, objections). In B2B you need both: first find the right company, then convince the right person inside it.

Do I need an ICP even as a small business?
Especially then. The tighter your time and budget, the more expensive every deal that does not fit. An ICP need not be elaborate – five criteria on one page are enough to triage enquiries faster and create focus where it counts most.

How much data do I need to derive an ICP?
Even 20–30 real customers are enough for a first pattern. More important than volume is honesty: rate by profitability and effort, not by how much you like them. Over time, and with more data, the profile becomes more precise and reliable.

How often should I review my ICP?
Schedule a review at least once a year, and every six months if you are growing fast. Markets, offers and competition shift; an ICP left untouched for two years often describes customers you no longer even want to win.

Can I have several ICPs?
Yes, but keep the number small. Two or three clearly separated profiles (for example by industry or size) are manageable. Maintaining five or more usually means no focus at all – at that point the tool has gone blunt.

What if a good enquiry does not fit the ICP?
The ICP is a guardrail, not a prison. Take the deal if it pays off – but note the deviation. If such cases pile up, that is a signal your ICP needs adjusting. Exceptions are data, not defeats.

How to tackle it – your checklist

Tick off these points and you have a usable ICP rather than a theory slide:

• Pulled a list of best customers from the last 2–3 years
• Marked the top 10–20% by profit and effort
• Analysed problem customers – noted exclusion criteria
• Defined 5–7 testable criteria (firmographic + behavioural)
• Set a simple points/weighting model
• Set A/B/C thresholds and the response to each
• Embedded the ICP score as a mandatory CRM field
• Put a review date in the calendar (every six to twelve months)

In the end an ICP is not a document but a habit: consciously triaging every enquiry. You can start free at advanzo.app – no credit card – and run your ICP as a field directly on your contacts. For agency setups, reach us at hey@advanzo.ch; more on that in CRM for agencies.