ICP Fit Scorer

Yes. Default weights are balanced across four ICP dimensions, but real lead scoring models almost always need tuning:

• Phase 1 (defaults) — treat the score as relative ranking, not an absolute threshold.
• Phase 2 (manual weights) — after 1-2 quarters of pipeline data, raise the weight of the dimension that best predicts close-won.
• Phase 3 (segment weights) — split mid-market vs enterprise, vertical A vs B; do not average them away.
• Phase 4 (backend weights) — move config to backend so weights can be A/B tested and rolled back.

Default weights are fine to start; value comes from running consistently and adjusting on real outcomes.

Yes. A shared ICP scoring model is one of the highest-leverage alignment tools — if both teams agree on the rubric and review it together.

• Single source of truth — one definition of "good lead"; no MQL vs SQL handoff fights.
• Shared rubric — write down what 0, 3, and 5 look like per dimension so scoring is consistent across reps.
• Quarterly calibration — review conversion by score band and adjust weights with data, not opinion.
• Override log — capture why reps override the score; that log becomes input for the next revision.

Score inflation usually means a rubric that is too generous, not a sudden surge of perfect-fit leads.

1. Tighten definitions — "3" should be common, "4" earned, "5" rare.
2. Add disqualifiers — wrong industry, sub-threshold size, locked into a competitor should cap or zero the score.
3. Recalibrate against won-lost data — re-score the last 50 deals; if wins and losses score the same, the model is not discriminating.
4. Shrink the high-score band — if 60% of leads land in Tier A, raise the cut to capture the top 15-25%.
5. Audit input quality — enrichment and form data often bias upward; sample manually.

No. The ICP scorer is a prioritization layer — discovery and qualification still need a human.

• Score handles ranking, removes triage subjectivity, and routes Tier A faster than Tier C.
• Score cannot read buying intent, champion strength, decision process, budget timing, or competitive dynamics.
• Together: route Tier A to a high-touch discovery call; route Tier C to automated nurture. Use the call to confirm or override the score.
• Feedback loop: when discovery contradicts the score, feed that signal back into the rubric.

Three tiers are enough for most teams. More tiers feel precise but add overhead without changing rep action:

• Tier A — senior AE, custom playbook, accelerated SLA. Typically 15-25% of inbound.
• Tier B — standard rep follow-up, structured nurture, qualification call. Typically 40-50%.
• Tier C — automated nurture, low-touch sequences, periodic re-scoring. The remaining 25-45%.

If you want four or five tiers, ask whether each adds a distinct action. Same AE + same playbook = a label, not a workflow.