ICP fit is how closely an account matches your ideal customer profile, and an ICP fit score turns that match into a number. The score is built from account attributes such as headcount, tech stack, and operating volume, not from buyer behavior. It tells you whether an account should buy, not whether it is ready to buy. Updated November 2026: added a weighting method from won and lost deals, a worked example, gates, and fit tiers.
Whether an account is ready to buy this quarter is a separate question, and blending the two into one number is where most scoring models lose their value.
What does ICP fit mean?
ICP fit is a measurement, and the ideal customer profile is the thing it measures against. The profile describes the companies that get the most value from the product and are most valuable in return. Fit says how close a given account sits to that description.
Two accounts in the same industry can have very different fit. One has 120 people in the function the product serves and runs the CRM it integrates with. The other has 15 people and a homegrown system. A target market slide lumps them together. A fit score separates them. If the profile itself is still a draft, start with how to build an ideal customer profile and come back to scoring once it is written down.
What goes into the score
Build the model backward from outcomes rather than from a target market slide. Take the accounts that closed, renewed, and expanded, then compare their attributes against accounts that lost, stalled, or churned. Most published ICP guides study winners only. Without the losers there is nothing to compare against, so every trait the winners share looks important. Any attribute that shows up at the same rate in both groups carries no information, however much the go-to-market team likes it.
Attributes that usually earn a place:
- Firmographics that gate the use case, such as headcount in the affected function or number of locations. - Technographics that determine whether the product can be implemented at all. - Operating characteristics that drive value, such as deal volume, transaction count, or sales team size.
Leave buying signals out. Website visits, content downloads, and trigger events measure timing, and they belong in a separate field (more on that below).
The data does not need to be clean. The objection ORM hears most in a deal is "our data is not ready." Pete Furseth, ORM's COO, answers it in one line: "Perfect data never happens. Predictive data is what matters." What matters is that each attribute is captured the same way across records. Consistent imperfect data still ranks accounts well.
How to calculate an ICP fit score
Weight each attribute by how strongly it separates won deals from lost deals, then add up the weights an account matches. The formula:
Fit score = sum of (attribute weight × match), where match is 1 for a full match, 0.5 for a partial match (for example, 30 to 49 people against a 50-person line), and 0 for a miss. The weights add up to 100.To set the weights, compare the win rate when an account matches each attribute against the win rate when it does not. The gap between the two is the attribute's information. Divide each gap by the sum of the gaps you keep, and multiply by 100.
The table is an illustrative build from 400 closed opportunities, won and lost:
| Attribute | Win rate when matched | Win rate when not matched | Gap (points) | Weight |
|---|---|---|---|---|
| 50 or more people in the function the product serves | 32% | 12% | 20 | 40 |
| Runs Salesforce (the product's required integration) | 30% | 14% | 16 | 32 |
| 1,000 or more deals a year | 26% | 17% | 9 | 18 |
| Sells in North America | 24% | 19% | 5 | 10 |
| In the target industry | 22% | 21% | 1 | 0 (dropped) |
Cut the total into three tiers. Tiers are what the field can act on. A score of 71 versus 68 is noise.
- Tier 1 (70 to 100): route to a named rep. - Tier 2 (40 to 69): work when Tier 1 is covered, or nurture. - Tier 3 (under 40): do not create pipeline without a strong reason.
Worked example (illustrative), using the weights above:
- Account A has 120 people in the function, runs Salesforce, closes 600 deals a year, and sells in North America: 40 + 32 + 0 + 10 = 82, Tier 1. - Account B has 35 people (a partial match, half of 40), runs HubSpot, closes 2,500 deals a year, and sells in North America: 20 + 0 + 18 + 10 = 48, Tier 2. - Account C has 15 people, runs Salesforce, closes 300 deals a year, and sells in Europe: 0 + 32 + 0 + 0 = 32, Tier 3, even though it sits in the target industry.
Some attributes should be gates, not weights. If the product cannot be implemented without Salesforce, any account that misses that row is out, whatever it scores. Under that rule Account B is disqualified before scoring. Gates belong in your written disqualification criteria, and keeping them out of the weighted sum stops a strong match elsewhere from pushing an unworkable account into Tier 1.
Then test the tiers on deals you did not use to set the weights. If Tier 1 does not win clearly more often than Tier 3, the weights are wrong. Sample size matters here. ORM looks for at least a couple of hundred closed-won deals before it trains its forecasting models, and Pete is blunt about companies below that line: "If you are closing a handful of deals a month, you don't have enough volume for the machine learning models." A company in that position should set weights by judgment, write down why, and recheck them as deals close.
Fit is not intent
A high fit score means an account looks like a customer. It says nothing about timing. Fit is stable for quarters at a time, while intent moves week to week, which is why the two belong in separate fields. Combined into one number, a low-fit account with a burst of website activity outranks a high-fit account that is quiet this month, and reps chase the wrong one. Account C above, with a research spike, is exactly the account a blended score would promote.
Pete's view of intent feeds is that buyers trust them too much: "It is not bulletproof. It is just one signal that could provide an opening for an initial conversation. It doesn't qualify an opportunity." His rough guess, which he says is an assumption rather than a measured figure, is that about one third of accounts showing intent are potentially real opportunities. Screening for fit first is what filters the rest. How to read the two signals together is covered in ICP fit vs intent data, and how fit sits beside behavioral scoring in ICP fit vs lead score.
Where the score changes decisions
The score earns its keep in four places. It routes inbound so the best-fit accounts reach a rep first. It sets territory and account list construction. It gives reps a defensible reason to disqualify early instead of working a deal that will grind for two quarters and close small. It also segments the pipeline for planning, because fit tiers close at different rates and at different sizes.
That last use connects fit scoring to the forecast. Low-fit accounts drag the win rate down, close for less than they were carried at, and churn faster once they are live, so they suppress net revenue retention as well. Pete describes a common pattern: a pipeline carrying an average deal size of $80,000 while closed-won deals average $40,000. Splitting pipeline and closed-won by fit tier is the first cut to make when that gap shows up.
Pipeline weighted without regard to fit assumes every dollar has the same probability behind it, which is rarely true across tiers. Pete makes the same point about stage weights: applying one weight to every deal fails when the deals come from different motions. Reporting coverage by fit tier shows whether the quarter depends on accounts that historically buy or on accounts that historically stall. Fit is one input to deal-level models too. Predictive deal scoring combines account attributes with how each opportunity is moving, which is where timing comes back in.
Keep the score current
Fit weights age. Pete names a changed market as the most common reason a forecast misses: the business or the market moved, and the model still runs on old assumptions. A fit model has the same weakness. After a pricing change, a new segment push, or a launch that opens a different buyer, rerun the win rate gaps on recent deals and reweight. If the gaps have not moved, leave the model alone.
Frequently Asked Questions
What does ICP fit mean?
ICP fit means how closely a company matches your ideal customer profile, the description of the accounts that get the most value from your product and are most valuable to you in return. A high-fit account looks like your best customers on attributes such as size, tech stack, and operating volume. Fit says nothing about timing. That is what intent data measures.
How do you calculate an ICP fit score?
Take a few hundred closed deals, won and lost. For each attribute, compare the win rate when an account matches against the win rate when it does not. Drop attributes with no real gap, weight the rest in proportion to their gap so the weights sum to 100, and score each account as the sum of the weights it matches. Then cut the total into three tiers.
What is the difference between an ICP fit score and a lead score?
An ICP fit score rates the account on static attributes such as industry, size, and tech stack. A lead score usually blends those attributes with behavior like page views and email replies. Keeping them separate is better, because a hot behavioral signal from a bad-fit account otherwise gets promoted into a deal that will not close.
Is an ICP the same as a target audience?
They are different. A target audience is the broad market a company sells into, often described by role or segment. An ICP is narrower: the specific type of company that gets the most value and is most valuable in return. Buyer personas sit inside the ICP and describe the people at those companies. The fit score measures accounts against the ICP, not against the wider audience.
How often should the score be rebuilt?
Refresh the underlying account data on a schedule and rebuild the weights when the win pattern moves, such as after a pricing change, a new segment push, or a product launch that opens a different buyer. A model trained on the customers a company won three years ago will keep routing reps toward the market it used to serve.
What if our CRM data is too messy to score accounts?
Messy data is the normal starting condition, and it rarely blocks a usable score. What breaks a model is inconsistency, where the same attribute is captured differently across records and periods. Consistent imperfect data still supports accurate ranking, so fix the capture rules first rather than waiting for a clean database.
Put these metrics to work
ORM builds custom revenue forecast models that turn concepts like icp fit score into prescriptive action for your team.
Schedule a Demo