What is ICP Scoring?

ICP scoring rates every account 0-100 on how closely it matches your ideal customer profile, and in Cleanlist AI an AI qualification costs 5 credits per person, with a row the agent cannot answer not charged at all. The fields the score reads are priced per result: 1 credit for a company record, 1 for a verified work email, 10 for a phone number, 11 for both. A new workspace opens with 14 days of Pro: 3 seats, 250 credits, which is 50 AI qualifications before any money changes hands. Most teams spray outreach evenly across a whole database, and scoring first is what stops that.

ICP scoring (Ideal Customer Profile scoring) is a systematic approach to evaluating and ranking leads based on their resemblance to your best-fit customer profile. Unlike traditional lead scoring, which often relies heavily on behavioral signals like email opens and page visits, ICP scoring emphasizes firmographic and technographic fit, characteristics that indicate whether a company is fundamentally a good match for your product, regardless of engagement activity.

What is an Ideal Customer Profile?

An ICP is typically defined by a combination of attributes: company size (revenue and headcount), industry or vertical, geographic location, technology stack, growth stage, and organizational structure. For example, a B2B SaaS company selling to mid-market might define their ICP as: "Software companies with 50-500 employees, $10M-$100M revenue, headquartered in North America, using Salesforce as their CRM, and currently in a growth or expansion stage."

ICP scoring assigns weighted values to each of these attributes and calculates a composite score for every lead or account in your database. A company that matches all criteria scores highest (e.g., 95/100), while one that only partially matches scores lower (e.g., 40/100). The weights reflect how predictive each attribute is of successful deals, if 80% of your best customers are in SaaS, then industry should carry more weight than, say, location.

How much does ICP scoring cost?

The score is a filter over data you already hold, so the bill is the data underneath it. In Cleanlist AI an AI qualification is 5 credits per person and writes the agent's reasoning next to the number, and a row the agent cannot answer is not charged. You spend only on the rows you decide to qualify. The firmographic and contact fields are priced per result: 1 credit for a company record, 1 for a verified work email, 10 for a phone number, 11 for both. A Starter seat is $49 a month with 750 credits, which is 150 AI qualifications. A Pro seat is $89 a month with 1,500 credits, which is 300. Annual billing takes 25% off seats. Full detail on pricing.

What data do you need before you can score an ICP?

A scoring model can only read fields that are actually populated, so enrichment comes first and scoring second. Cleanlist AI returns industry, employee count, HQ location, company domain and founded year on the company side, and job title, seniority, department and LinkedIn URL on the person side, all normalized on the way in. Cleanlist AI does not sell a revenue database, a technographic feed or an intent feed. Where a rubric needs revenue scale, tech stack or a growth signal, an AI column researches each account on the open web and writes back what it found along with its reasoning, which is a different thing from a purchased dataset and should be weighted accordingly. Score on what you can verify, and treat researched signals as tiebreakers.

ICP scoring vs behavioral lead scoring

The advantage of ICP scoring over behavioral lead scoring is that it identifies good-fit companies even before they engage with your marketing. A perfect-fit company that has never visited your website is still a valuable prospect, while a poor-fit company that downloads every whitepaper is still unlikely to convert.

The best qualification systems combine both ICP scoring (fit) and behavioral scoring (intent) for a complete picture. A high-fit, high-intent account should be routed immediately to a senior AE. A high-fit, low-intent account belongs in nurture sequences. A low-fit, high-intent account can be deprioritized despite its engagement.

How to build an ICP scoring model

Step 1: Analyze your best customers. Look at your top 20-30 accounts by revenue, retention, and expansion. Identify common firmographic patterns, are they concentrated in specific industries, company sizes, or geographies?

Step 2: Define scoring attributes and weights. Select 5-8 attributes that correlate with success. Assign weights based on their predictive power. A common starting framework:

• ●Industry match: 25 points (exact match) / 10 points (adjacent)
• ●Revenue range: 20 points (ideal range) / 10 points (close)
• ●Employee count: 15 points (ideal range) / 5 points (close)
• ●Technology stack: 15 points (uses key technologies)
• ●Geography: 10 points (target market)
• ●Growth signals: 10 points (hiring, funding, expansion)
• ●Negative signals: -15 points (wrong industry, too small, etc.)

Step 3: Score and validate. Apply the model to your existing customer base and pipeline. If the model doesn't correctly rank your best customers above your worst, adjust the weights. A well-calibrated model should show clear separation between won and lost deals.

Step 4: Iterate based on results. Review model accuracy quarterly. As your product evolves and your market understanding deepens, update the criteria and weights. Track the conversion rate of each score tier (90+, 70-89, 50-69, below 50) to validate that higher scores actually predict better outcomes.

ICP scoring criteria for B2B sales

ICP scoring criteria are the weighted firmographic, technographic, and behavioral attributes that determine how well a prospect matches your ideal customer profile. For B2B sales teams, a clear set of criteria turns a vague "good fit" gut call into a repeatable 0-100 score that any rep can trust. The criteria split into three signal categories, each answering a different question about the account.

Firmographic signals (who the company is): industry or vertical, annual revenue, employee headcount, geographic location, growth stage, and funding status. These describe the company's fundamental shape and rarely change month to month.

Technographic signals (what the company runs): CRM platform, marketing automation stack, cloud provider, and any complementary or competing tools that signal readiness to buy. A company already running Salesforce and a modern sales-engagement tool is a stronger fit for most B2B SaaS products than one with no detectable stack. Cleanlist has no technographic database behind this category. An AI column researches the open web per account and reports what it found, so the evidence is visible and the weighting is yours to set.

Behavioral signals (what the company is doing): hiring for relevant roles, recent funding, website visits, content downloads, and product-page engagement. These are dynamic and best layered on top of fit, not used as the primary qualifier.

ICP scoring rubric: a B2B SaaS definition and example

An ICP scoring rubric is a documented table that assigns point values to each criterion so every account receives a consistent, defensible score. Here is a concrete b2b SaaS ICP scoring rubric definition that a mid-market sales team can use as a starting point:

An account scoring 80-100 is a priority-fit account routed straight to an AE. 60-79 enters SDR-led nurture. Below 60 is deprioritized. The weights are not universal: a team selling enterprise security weights tech stack and revenue higher, while a product-led startup weights growth signals higher. The rubric is a living artifact, recalibrated quarterly against which score tiers actually convert.

This rubric-based approach is how teams operationalize buyer identification at scale. Tools like Clay and 6sense let you assemble similar target buyer identification workflows by enriching accounts and scoring them against criteria, but they require you to build and maintain the rubric and the enrichment plumbing yourself. Cleanlist AI applies the rubric automatically as records are enriched, so accounts arrive in your CRM already scored against your ICP scoring criteria.

Why enrichment has to happen before scoring

A scoring model is only as good as the fields it can read, and an empty column silently scores as a zero, quietly demoting a good account. Three gaps break models in practice. Stale industry classifications, because companies pivot, merge and reclassify, and last year's code produces this year's wrong score. Missing headcount, which is the single field most rubrics lean on hardest. And criteria written against data you do not hold, which is the most common failure of the three: a rubric weighting revenue at 20 points against a database with no revenue column has quietly switched 20 points off.

Cleanlist AI closes the first two by enriching the record before it is scored, routing each row through 25+ providers and normalizing industry, employee count, HQ, domain, founded year, title, seniority and department on the way in. On the Cleanlist AI 500-Lead Enrichment Benchmark, 2026, that cascade returned a verified email for 98% of 500 stratified B2B leads and a phone number for 85%, against 70-80% and 30-60% from single-source databases on the identical leads. The third gap is a design decision rather than a data problem: write the rubric against fields you can actually populate, and let an AI column research the softer signals so the reasoning is visible per row. Many sales prospecting tools now include some form of ICP scoring, see the full comparison at best sales prospecting tools for how different platforms approach lead qualification.