ICP refinement in 90 minutes

Why most ICPs are too loose

The typical ICP doc says 'mid-market SaaS, 50–500 employees, US-based, decision-makers in sales or marketing.' That's not an ICP — that's a vague aspiration.

A real ICP has specifics that translate directly into Sales Navigator filters. 'B2B SaaS in the developer tools or DevOps space, 80–400 employees, headquartered in US/UK/Canada, currently hiring 2+ engineers, raised Series B or later, founded after 2018, technical co-founder still actively involved.'

The second one identifies maybe 600 companies. The first one identifies 60,000. Outbound to 60,000 is spam; outbound to 600 is craft.

Step 1 — Inventory your best customers (15 min)

List your 5–10 best current customers. 'Best' = highest LTV, fastest sales cycle, lowest churn risk, most enthusiastic referrers. Not biggest revenue — biggest fit.

For each, write down: industry / sub-industry, company size (employees and revenue if you know), tech stack (what tools do they use that matter to your product), why they bought (the trigger event or pain), how long the sales cycle was, who the buying committee was.

By the end of step 1 you should have a 5–10 row table you can stare at.

Step 2 — Find the patterns (20 min)

Pattern-match across the table. What's common? What's different? Three patterns usually emerge:

Firmographic patterns. Industry concentration, size band, geography, founding age.

Technographic patterns. They all use Tool X. They all moved off Tool Y in the last year. They all have an internal team running Function Z.

Trigger patterns. They all bought within 6 months of [event]. New leadership in role X. Recent funding. New product launch. Regulatory change.

Write down each pattern as a one-line claim. By the end of step 2 you have 6–10 specific claims.

Step 3 — Define disqualifiers (15 min)

Equally important to who you DO target: who you DON'T target. Disqualifiers prevent the wasted outreach that dilutes reply rates.

Common disqualifiers: company size below X (won't have budget), above Y (won't move fast enough), industry Z (regulated and won't use your product), tech stack mismatch, geography mismatch, recent restructuring (won't make decisions).

Aim for 4–6 explicit disqualifiers. Each one should have a specific reason — not just 'we don't want them' but 'they won't close because [X].'

Step 4 — Translate to Sales Navigator filters (25 min)

Convert each pattern and disqualifier into a Sales Navigator filter combination. This is where the rubber meets the road.

Industry: primary industry filter + sub-industry keyword filter + disqualifier exclusions.

Size: employee range filter + revenue range if available.

Geography: country / region / metro filter.

Tech stack: Sales Nav doesn't have this directly — use ZoomInfo or Apollo to layer technographic data.

Trigger events: 'Recent funding' filter, 'Hiring spike' (job posting count), 'New leadership' (recent role changes filter).

Save 3–5 distinct named searches in Sales Navigator. Don't try to capture everything in one search — multiple searches with overlapping but distinct criteria capture more of the actual pool.

Boolean search cheat sheet for the exact filter syntax.

Step 5 — Define the buying committee (15 min)

Inside each target account, who actually decides? Most B2B deals have 3–6 person buying committees: economic buyer, champion, technical evaluator, finance approver, end user.

For each role in the committee, specify the LinkedIn-searchable title patterns. 'Economic buyer = VP Sales OR Chief Revenue Officer OR Head of Revenue OR Founder/CEO at companies under 50 people.'

Different roles need different first messages. The economic buyer cares about ROI; the technical evaluator cares about how it integrates with their stack; the end user cares about whether their daily life gets better.

By the end of step 5, you have 3–5 personas with different LinkedIn search patterns and different first-message angles.

What good output looks like

After 90 minutes you should have a one-page document with:

• 6–10 specific firmographic / technographic / trigger patterns

• 4–6 explicit disqualifiers

• 3–5 named Sales Navigator searches saved and active

• 3–5 buyer-committee personas with title patterns

• A clear sense of how many target accounts this represents (use Sales Nav's count to estimate)

If your final ICP document doesn't fit on one page, it's still too vague. Cut.

FAQ

What if I don't have customers yet?

Use proxy customers — companies that look like the customers you'd want. The pattern-matching is the same. Validate with cold outreach: target 30 companies that match your ICP, see who responds and why.

How often should I refresh the ICP?

Quarterly. Patterns change as your product evolves and the market shifts. The disqualifiers especially get refined fast in the first year.

Should I have one ICP or multiple?

Multiple is fine, especially if you serve distinct segments. Don't have more than 3-4 — beyond that you're spreading too thin to do any of them well.

What about expansion vs new-logo ICPs?

Different exercises. This guide is for new-logo ICP. Expansion ICP works differently — pattern-match across customers who upgraded vs churned.