Docs

ICP best practices

What makes an ICP that actually finds great leads — and the mistakes to avoid.

5 min read

A precise ICP returns fewer, better leads. A vague ICP returns many noisy leads. Use these patterns to keep yours sharp — and the next section for the traps to avoid.

Do this

  • Start from your best customers. Open your CRM, list your five or ten happiest accounts, and find the patterns. Industry, size, geography, the buying signal that mattered. That's your ICP.
  • Be specific about company size. A 5–10× range (50–500 employees, $5M–$50M revenue) is sharper than "SMB". Specific bands let the AI score on a precise signal.
  • List exclusions explicitly. Tell Aidealy what you don't want — agencies, non-profits, certain geographies. Exclusions clean the list as effectively as inclusions sharpen it.
  • Iterate weekly. Run your ICP, scan the top scorers, ask yourself "would I want to talk to this company?". If the answer is no for a pattern of leads, tighten the ICP. Iteration beats analysis every time.
  • One ICP per motion. If you sell to two very different customer types, build two ICPs. Trying to fit both into one set of criteria produces mediocre matches for both.

Use signals to break ties

Buying signals — recent funding, hiring momentum, tech-stack changes — are tie-breakers between two otherwise-similar companies. Add the signals that genuinely correlate with your wins, not every signal you can think of. Quality over quantity.

Avoid this

  • Don't define your ICP by what you want to sell. Define it by who has the problem you solve. "Companies that need our product" is circular; "B2B SaaS companies with 50+ sales reps" is concrete.
  • Don't pick more than three industries. Beyond three, you're no longer targeting — you're scattering. Build separate ICPs if you need to cover more.
  • Don't use vague geographies. "Global" is not a geography. Pick the regions where you can actually close deals, even if it's a long list.
  • Don't set the company size too wide. "10–10,000 employees" treats a five-person startup the same as a Fortune 500. The signals matter for both, but the matches won't.
  • Don't forget to refresh. An ICP that worked last year might be missing today's signals. Revisit every couple of quarters.

Watch out for over-fitting

It's tempting to build an ICP that perfectly describes your last big win. Resist — one customer is anecdote, not pattern. Use at least five recent wins as the basis, and look for what they share, not what makes any single one special.

Frequently asked

One per distinct sales motion. Most teams end up with 1–3: maybe a primary ICP for your main product and a secondary one for a different segment or pricing tier. More than five is usually a sign you're slicing too thin.