There’s a version of ICP work that everybody does. You look at your best customers, find the common firmographic attributes (industry, company size, geography, tech stack), and write them down. Then you use those attributes to filter leads from a database. This works. It gets you in the right neighborhood.

But it doesn’t get you into the right house. The companies that match your firmographic filters but never convert (and there are always a lot of them) are proof that something is missing from the definition.

The “deep ICP” is about finding what that missing layer is.

Surface vs. depth

Firmographic attributes are surface-level descriptors. They tell you what a company looks like from the outside. They don’t tell you why a company buys.

The reasons companies buy tend to involve things you can’t find in a database dropdown:

• Organizational pain. They just lost a key person. They missed a compliance deadline. Their current process broke during a growth spurt.
• Strategic context. They’re entering a new market. They’ve just raised funding. They’re consolidating vendors.
• Operational signals. They’re hiring for a specific role. They posted a job description that mentions the exact problem you solve. Their tech job listings changed from one stack to another.
• Behavioral signals. They attended a competitor’s webinar. They’re active in a relevant community. They published a blog post about the problem space.

These are harder to filter for, but they’re often more predictive than firmographics.

Building the deep ICP

Here’s a process that works:

Step 1: Interview your closed-won deals

Go back to your last 10–15 closed deals. Not the ones that looked good on paper, the ones that actually closed. For each one, ask:

• What was happening in the company when they started evaluating?
• What was the triggering event?
• Who initiated the search?
• What alternatives did they consider?
• Why did they choose you?

You’re looking for patterns that go deeper than “they’re a 100-person SaaS company.” You might find that 80% of your deals were triggered by a specific event (new hire, failed audit, lost a big deal because of a process gap). That event is part of your deep ICP.

Step 2: Define the signal questions

Take the patterns from Step 1 and turn them into questions you can evaluate from publicly available information:

• “Has this company posted a job for [specific role] in the last 60 days?”
• “Does their website mention [specific compliance framework]?”
• “Have they recently announced [event type: funding, expansion, acquisition]?”
• “Do they have a team or function that would own [your product category]?”

These are your qualifiers. They’re more useful than firmographic filters because they capture the context in which a company becomes a buyer, not just the category the company belongs to.

Step 3: Weight by predictiveness

Not all signals are equally useful. You might find that “recently hired a Head of Revenue Operations” is a very strong predictor, while “B2B company in the UK” is necessary but not sufficient. The qualifier approach lets you combine these with different weights or at different tiers.

A common structure:

• Hard requirements (must pass all): Industry, geography, company type. These eliminate obviously wrong leads quickly.
• Signal qualifiers (scored): Recent relevant hire, specific tech in use, published content about the problem. These separate “matches the profile” from “is actively in a buying context.”

Step 4: Test against reality

Here’s where most ICP work stalls: the definition gets written but never tested. The deep ICP only works if you validate it against actual outcomes.

Run your qualifier set against a batch of leads. Qualify them. Then track what happens:

• Of the leads that passed, how many engaged?
• Of the leads that failed, were there any that would have been good prospects?
• Which qualifiers had the most false positives or false negatives?

Adjust. This is an iterative process. The first version of your deep ICP will be wrong in interesting ways. The third version will be notably better.

Why bother going deeper?

Smaller, better lists. When you qualify on context and signals, not just firmographics, you discard the leads that look right but aren’t. Your list gets shorter and your conversion rate goes up.

Better messaging. When you know why a company is in-market, not just that they fit a demographic profile, you can write outreach that speaks to their specific situation. “I noticed you just hired a Head of Data Privacy” beats “I see you’re a mid-market SaaS company” every time.

Evidence for compliance. In GDPR jurisdictions, demonstrating that you had a specific, documented reason to contact someone is central to a legitimate interest claim. A deep ICP with explicit qualifiers produces exactly this kind of evidence. (See: Legitimate Interest Assessments for Outbound.)

Fewer wasted cycles. Your BDRs stop burning hours on accounts that look right on paper but were never going to buy. That time goes back into the accounts that actually matter.

Putting the deep ICP to work

The challenge with the deep ICP is that it requires evaluation, not just filtering. You can’t set a dropdown to “recently hired a DPO.” You have to actually look.

This is where AI evaluation helps. In BDR², you define qualifier questions, and the system evaluates them against each lead using AI agents. The outputs feed into an ICQ mask (a logical rule), and you get a pass/fail with a full evidence trail.

But even without tooling, the framework applies. If you’re doing outbound research manually, write down the qualifier questions and answer them for each lead before deciding whether to reach out. Asking the same questions consistently will improve your targeting fast.

Keep updating it

The deep ICP should not be a static document. The signals that predict buying change as your product evolves and as you learn from outcomes. Version your qualifiers. Track which ones predict conversions. Retire the ones that don’t. Add new ones when you notice new patterns.

Good outbound comes from knowing, in detail, why your customers buy. The deep ICP is how you build that knowledge.