A focused ICP outperforms a broad one. Define it with concrete firmographics, add technographics only when they predict fit, always list disqualifiers, and revisit after 30 days with real deployment data. Narrow beats aspirational.
Start narrow, then widen
Write the ICP for the segment you already win in, not the total addressable market. A rubric that surfaces 2,000 excellent-fit accounts beats one that surfaces 40,000 mediocre ones — the agent's budget is finite, so precision compounds. You can always add adjacent segments once the core is producing.
Use disqualifiers aggressively
Every rubric should list segments to never pursue: wrong company size, competitors, existing customers, unsupported regions or languages. Disqualifiers prevent wasted credits and embarrassing outreach, and they're often more valuable than the positive criteria.
Weight by predictive value, not intuition
When you set criterion weights, anchor on what actually correlated with closed-won deals — not what feels important. If industry predicts fit far more than headcount, weight it accordingly. Review the score breakdown on a few known-good and known-bad accounts to sanity-check the weights.
Iterate on real data
After 30 days, compare predicted tier against who actually engaged and converted (see Tune your ICP after 30 days). Promote criteria that predicted well, drop those that didn't, and adjust tier thresholds so A-tier stays a genuinely scarce, high-intent set.
