What Is an Ideal Customer Profile (ICP)? Definition, Examples, How to Build One

The one-sentence definition

An Ideal Customer Profile (ICP) is a written description of the type of company that, when sold to, gets the most value from your product the fastest, the cheapest, and is the most likely to renew or refer. It is not a buyer persona — that's a person. ICP is the company.

What an ICP is not

• Not a wish-list. "Fortune 500 companies" is a fantasy, not an ICP.
• Not a buyer persona. The persona ("VP of Engineering, 35–50, has used Datadog") sits inside the ICP.
• Not your TAM. Total addressable market is "everyone who could plausibly buy". ICP is "who you should target first".

The five fields every ICP must have

1. Firmographics — industry, sub-industry, company size (employees, revenue), geography, public/private.
2. Technographics — what tools they already use that signal fit. Selling email infra? Targets that already use SendGrid are warmer than targets that send through Gmail.
3. Triggering events — recent funding round, leadership change, new office, hiring spree, security incident, regulatory deadline. Triggers beat demographics.
4. Pain you can name in one sentence — "Their support queue gets backed up over the weekend because no human is on" beats "they need better support".
5. Disqualifiers — equally important. Who looks like a fit but isn't? (Wrong region, regulatory blocker, can't afford the floor price.)

A worked B2B SaaS example

Product: AI customer support copilot.

ICP: US/EU-based mid-market SaaS companies (50–500 employees, $5–50M ARR) that run support on Zendesk or Intercom, have a team of 6–30 support agents, support a product with a self-serve free tier (so support volume is high), and have raised a Series B in the last 18 months (budget unlocked).

Disqualifiers: healthcare (HIPAA blocker), Asia-Pacific headquartered (timezone friction with our team), Salesforce Service Cloud users (different stack, not our integration target).

A worked freelance/agency example

Product: Done-for-you website redesign for local businesses.

ICP: Independently-owned restaurants and dental clinics in tier-1 Indian cities, currently rated 4.3+ on Google Maps with 200+ reviews (signals: real, established, has revenue), with a website that scores below 70 on Lighthouse mobile (signal: clear gap), no active Instagram link from the site (signal: low marketing maturity).

Disqualifiers: Chains with 4+ locations (centralised marketing — our pitch lands wrong), already using Wix/Squarespace with a recent template (already addressed the pain).

The second example is exactly the kind of ICP that Cold Scout's Maps-based discovery is built for.

How to actually build yours

1. List your last 20 closed-won customers. Strip outliers (the favor deal, the pilot that didn't pay).
2. For each, write down the five fields above as they actually were when the deal closed.
3. Look for the cluster. The ICP is the description that fits at least 60% of the rows.
4. Validate against your last 10 closed-lost. If the lost deals match the same description, your ICP isn't tight enough — add a discriminator.

ICP scoring (the fun part)

Once the ICP is written, encode it as a scoring rubric so you can rank inbound and outbound leads automatically. A simple weighted formula:

score = 0
score += 30 if industry matches
score += 20 if company size in [50, 500]
score += 20 if geography matches
score += 15 if uses one of {tool_a, tool_b}
score += 15 if has triggering event in last 90d
score -= 50 if any disqualifier

That's it. A 0–100 score, every lead, every day, no judgment calls. Cold Scout does this for Maps-sourced businesses by passing the ICP rubric to a small Llama model and asking it to score and explain — same idea, more inputs.

When to revisit

Every quarter at minimum, after a major product change, or whenever your win rate drops on a specific segment. ICPs aren't static — yours should evolve as your product does.