Short answer
Pick five or six criteria that separate your best customers from the rest, such as industry, size, region, customer segment, tech stack and intent. Weight them to 100 points, score full, partial or no match, and set A/B/C thresholds. Then back-test: score your last 30 won and 30 lost deals. If most wins don't land in A and B, change the weights before anyone uses the score. Keep fit and intent visible as separate numbers.
Choose the criteria
| Criterion | Example of a full match | Starting weight |
|---|---|---|
| Industry | Tier 1 industries where you win most | 25 |
| Company size | 200–2,000 employees | 20 |
| Intent signals | Research on your category or a trigger event | 20 |
| Customer segment | Sells to the same customers as your best accounts | 15 |
| Region | Core markets with sales coverage | 10 |
| Tech stack | Uses systems you integrate with | 10 |
Score each criterion as full, partial or no match, and add the points. The ICP scoring calculator scales any weights to 100 and exports the model.
Set A/B/C thresholds
- A (70+): candidates for 1:1 or 1:few plays and LinkedIn committee targeting.
- B (50–69): 1:many programs and AI ads follow-up.
- C (under 50): no paid ABM unless intent is strong.
Thresholds should produce a manageable A list, often 5–15% of accounts. If half the market scores A, the model isn't selective enough.
Back-test before rollout
- Score the last 30 won and 30 lost or stalled deals as they looked at the start.
- Check the share of wins in A and B. You want most of them there, and losses spread lower.
- Adjust the weight of any criterion that wins and losses share equally; it isn't telling you anything.
- Repeat once, then freeze the model for six months.
Keep fit and intent apart
Only a small share of accounts is in market at once; the B2B Institute's 95:5 rule puts it near 5%. Fit says who should be on your list all year. Intent says who to push this week. Show both on the account record, so a high-fit account without intent stays in always-on programs and an intent spike on a C account doesn't jump the queue.
Roll it into the CRM
- Store the score and each criterion's points as fields, so sales can see why an account scored as it did.
- Recalculate weekly for intent and monthly for firmographics.
- Use the grade to build LinkedIn and DSP lists, and to tier the budget with the ABM tiering calculator.
- Score companies identified from AI traffic the same way. How.
Questions
What is an ICP fit score?
A number, usually 0–100, showing how closely an account matches your ideal customer profile, built from weighted firmographic and technographic criteria.
Should intent be in the ICP score?
Keep it visible separately. Fit changes slowly and tells you who to target; intent changes weekly and tells you when.
How many criteria should an ICP model have?
Five or six. More criteria rarely change the ranking and make the score harder to explain to sales.
How often should we revise the model?
Every six to twelve months, or after a big product or market change, using fresh win and loss data.
Facts checked against the sources below on 19 October 2026. Ad platforms change quickly: if something here no longer matches what you see, tell us and we will update it.
