How to Rank Companies Against an ICP With Match Scores and Source Evidence

You rank companies against an ICP by scoring each account against a fixed set of weighted criteria (firmographic, technographic, and behavioral), then attaching the specific piece of evidence, a job posting, a detected tool, a funding filing, that justifies every point, so the resulting list is ordered by fit and every score can be checked rather than just trusted.

This guide is for founders, sales leaders, RevOps, and SDRs who want a repeatable way to turn an ideal customer profile into a ranked, workable list, without ending up with a score nobody on the team believes. It covers what goes into the score, why the evidence behind it matters as much as the number, and how to keep the ranking current as company data changes.

How do you rank companies against an ICP with match scores?

You start by writing down what your ICP actually is in concrete, checkable terms, not "mid-market SaaS companies" but specific ranges and signals: employee count between two numbers, industry, tech stack, hiring activity, funding stage, headquarters region. Each of these becomes a scoring criterion.

The method has four steps:

1. Define criteria and weights. Decide which firmographic, technographic, and behavioral factors predict a good customer for you, and how much each one should count toward the total. Not every criterion deserves equal weight: a company using a competing or complementary tool in your stack usually matters more than its exact headcount.
2. Pull the underlying data per company. For each account, gather the facts behind each criterion: company size, industry classification, detected technologies, recent hiring, recent funding, leadership changes.
3. Score each criterion and sum to a total. A common approach is 0 to 100, with each criterion contributing points up to its weight if the company matches, partial points for a partial match, and zero if the data does not support it or is missing.
4. Sort the list by score. The highest-scoring accounts go to the top of the list your reps work first.

The part teams usually skip is keeping a record of exactly what data justified each point. Without that, the score is a number you either trust blindly or ignore.

What data goes into an ICP match score?

Three categories cover most of what a useful ICP score needs.

Firmographic data answers the basic eligibility question: does this company even look like your customer. Employee count, industry, revenue range, and headquarters location fall here. This is the cheapest data to get and the least predictive on its own, since two companies with identical firmographics can have completely different buying readiness.

Technographic data tells you what tools a company already runs. If your product integrates with or replaces a specific platform, knowing that a company uses it is a strong fit signal. This is also detectable from public sources: job postings that name the tool, changelogs, trust center pages, or community activity. Our guide on how to find companies that use Salesforce walks through exactly how technographic detection works in practice, since Salesforce usage is one of the more common technographic criteria in B2B scoring.

Behavioral and signal data is the most time-sensitive category: hiring for specific roles, funding rounds, leadership changes, and public statements about priorities. A company that is actively hiring for roles that indicate a new initiative is showing intent that a static firmographic profile cannot capture. This is also the data that goes stale fastest, which is why a score built once and never refreshed drifts away from reality within weeks.

A simple weighting example:

The exact split depends on what actually correlates with your closed-won accounts, not a generic template. A company selling infrastructure tooling will weight technographic fit heavily; a company selling to a specific function will weight the hiring signal heavier.

Why does a match score need source evidence attached?

A score without evidence is a claim you cannot verify. If a rep sees "87" next to a company name and nothing else, they have two options: trust it or ignore it, and neither is a good use of a scoring system you spent time building.

Attaching evidence to each point solves a specific problem: some of this data is inferred, not looked up. A tool that classifies "industry" from a company description, or infers "growth stage" from a headcount trend, is making a judgment call, and judgment calls are sometimes wrong. When each contributing point links back to the specific fact behind it (this job posting, this detected integration, this funding announcement), a rep can check the two or three points that look off instead of re-verifying the whole account from scratch, and you can catch a systematically wrong criterion before it drags down the rest of the list.

Evidence also matters for maintenance. When a criterion's underlying source disappears (the job posting gets taken down, the tool gets replaced), a score with linked evidence can flag that the point is no longer supported. A score that was calculated once and stored as a static number cannot tell you that.

How do you keep the ranking accurate as it changes over time?

An ICP score calculated once is a snapshot, and snapshots age. A company that scored high because it was hiring for a target role three months ago may have already filled that role and moved on. Three practices keep a ranking useful past the first pass:

• Recompute on a schedule, not just at list-building time, so a company's position reflects current signals rather than the day you first scored it.
• Re-rank when new evidence appears, rather than waiting for a full refresh cycle. A company that just raised a funding round or opened a relevant role should move up the list the same week that becomes public, not at the next quarterly review.
• Let old evidence expire. A hiring signal from eight months ago is not the same strength of evidence as one from last week. Weighting recency into the score (or simply dropping stale signals) keeps the top of the list meaningful.

None of this replaces a qualification call. A high match score means an account is worth a rep's time to investigate, not that it is ready to buy. Treat the score as a way to order the work, not a substitute for doing it.

How does Retriever help here?

Retriever is built around exactly this workflow. You describe your ICP in plain English instead of assembling boolean filters, and Retriever returns a ranked list of matching companies, each with a match score and the specific source evidence behind it, so you can see why a company ranked where it did instead of taking the number on faith.

Underneath the score, Retriever sets up and continuously tracks the behavioral layer that keeps a ranking current: hiring activity, funding events, tech-stack changes, leadership moves, plus LinkedIn and social listening. A company surfaces or moves up the list when it fits the ICP and a relevant signal fires, instead of waiting for a manual re-scoring pass. For teams that want this running inside their own tooling or an AI agent's workflow, Retriever's MCP server exposes the same ICP search and scoring programmatically.

See how the scoring and evidence look in practice on the product page, or book a demo to run it against your own ICP.

Frequently asked questions

What is a good ICP match score? There is no universal cutoff, since the score depends on the criteria and weights you chose. What matters more is using the score relatively: work the top band first, treat the middle band as needing more signal before committing rep time, and periodically check whether the accounts that actually closed scored where you expected.

Can you build ICP scoring in a spreadsheet? Yes, at small volumes. Define your criteria and weights in columns, fill in the data manually or through a CSV export, and sum the rows. It becomes hard to sustain once you need to refresh dozens of data points across hundreds of accounts on a recurring basis, which is where the manual approach usually breaks down first.

What's the difference between lead scoring and ICP scoring? Lead scoring typically ranks individual people based on engagement (email opens, form fills, page visits). ICP scoring ranks companies based on fit criteria, independent of whether anyone at that company has engaged with you yet. The two are complementary: ICP scoring tells you which accounts to target, lead scoring tells you which contacts within those accounts are warming up.

Does a high match score mean a company is ready to buy? No. It means the company fits your ideal customer profile well enough to be worth prioritizing. Timing, budget, and internal appetite still have to be confirmed through actual conversation. Treat the score as a prioritization tool, not a buying signal on its own.

How often should ICP scores be recalculated? It depends on how fast your signals change. Firmographic criteria (size, industry) change slowly and can be refreshed monthly. Behavioral signals like hiring and funding change quickly enough that a weekly refresh, or event-driven updates when new evidence appears, keeps the top of the list from going stale.