June 2026·7 min read·Prospecting

How to Build an Ideal Customer Profile That AI Can Actually Use

“Mid-market B2B SaaS companies” is not an ICP. It's a vague hope. Here's a practical, 4-dimension framework for defining who you actually want to prospect — specific enough that AI can find them.

Why most ICPs produce bad results

Most companies have an ICP that looks like this: “We sell to B2B SaaS companies with 50-500 employees in North America.” This is not a profile. It's a starting point so broad it produces hundreds of results, most of which aren't actually good fits.

A useful ICP doesn't just describe who might buy. It describes who actually buys — specifically enough that a search tool or AI can distinguish a great fit from an okay one without human judgement on every result.

The difference between “B2B SaaS companies” and “B2B SaaS companies with 50-200 employees, a dedicated VP of Sales, and at least 3 job openings in the last 30 days” is the difference between 300 unqualified results and 30 that are genuinely worth prospecting.


The 4-dimension ICP framework

An ICP that produces good results needs four dimensions. Most companies define the first and stop. The other three are where the precision comes from.

1

Firmographics

Industry, company size, location, revenue. The basics. Necessary but not sufficient.

2

Signals

Behaviours that indicate need: recent funding, hiring patterns, technology stack, growth stage. This is where ICPs go from broad to precise.

3

Decision-makers

Who you need to reach: job titles, seniority, function. Not just who has budget — who feels the problem you solve.

4

Exclusions

Who looks like a fit but isn't. The companies you don't want to waste time on. Equally important as inclusions.


Dimension 1: Firmographics (the basics)

These are table stakes. Get them right and you narrow from “every company” to “companies in your market.” But don't stop here.

AttributeVague (Bad)Specific (Good)
Industry“Technology”“SaaS companies selling to HR teams”
Company size“SMB”“20-200 employees”
Location“UK”“London, Manchester, or Birmingham metro areas”
Revenue“Established”“£2M-£20M annual revenue”

Dimension 2: Signals (where ICPs go from broad to precise)

Signals are observable behaviours that indicate a company is likely to need what you sell. They're the difference between “companies that could buy” and “companies that are probably looking right now.”

Hiring patterns

Hiring for a role your product replaces or supports. If you sell dev tools, companies hiring engineers faster than average.

Funding events

Recently raised a round. Money in the bank and pressure to grow. Timing window: 3-6 months post-raise.

Technology stack

Using specific tools that integrate with or compete with yours. If you sell a Salesforce add-on, companies using Salesforce.

Growth indicators

Rapid headcount growth, new office openings, expanding into new markets. Companies in scaling mode.

Content & activity

Active blog, job postings, social presence. Companies that are “alive” vs. dormant LinkedIn profiles from 2019.


Dimension 3: Decision-makers

Don't just target the obvious title. Map the org chart of who feels the problem, who champions the solution, and who signs the cheque. Often they're different people.

The Champion vs. the Buyer

In a 50-person company, the Head of Operations feels the inefficiency you solve. The CEO signs off on new tools. You need to reach the champion (to build internal momentum) and the buyer (to close). Your ICP should include both.

Dimension 4: Exclusions (who you don't want)

Exclusions are as important as inclusions. They prevent you from wasting time on companies that look right on paper but never convert. Define them explicitly.

×

Companies in your ICP industry but with the wrong business model (e.g., marketplace vs. SaaS)

×

Companies at the right size but with no dedicated function for what you solve

×

Companies in the right location but serving a market you don't understand

×

Companies that would need 6+ months of onboarding before seeing value (mismatched expectations)