How to Use Clay for Account Scoring: A Practical Guide

Account scoring usually breaks at the data layer. Half your CRM records have gaps, so the model outputs garbage tiers. Reps lose trust, the project gets shelved, and your team is back to gut-feel prioritization within a quarter. Clay solves the data problem first, then lets you build scoring logic on top of clean records.

This blog walks through the full workflow: enrichment sequence, formula syntax, the signals that matter for $20K+ ACV deals, and the mistakes that quietly burn credits without producing usable output.

Why Clay works for account scoring

Clay connects to 150+ enrichment providers, gives you a spreadsheet where you write your own scoring formulas, and includes an AI research agent called Claygent. It pulls firmographic data from Apollo and Clearbit, technographic data from BuiltWith and HG Insights, and consolidates these sources inside a single table where you define your own logic.

The core mechanism is waterfall enrichment. Clay queries multiple data providers sequentially until it finds a match. Coverage gains have been reported to roughly triple enrichment coverage. Records without enrichment data cannot be scored, so coverage comes first.

The market has three tiers of scoring options:

• CRM-native scoring like HubSpot and Salesforce Einstein primarily uses CRM data. Salesforce Einstein can fall back on a global model built from anonymous cross-customer data when local conversion data is thin.
• Enterprise intent platforms like 6sense and Demandbase typically cost thousands of dollars per year for many deployments. Reviewers frequently describe them as opaque "black box" systems where the scoring methodology is not fully transparent.
• Clay sits between those two, with self-serve plans starting around $167 per month and custom enterprise pricing.

For Series A through C companies with a defined ICP, Clay offers full transparency, deep customization, and enough enrichment depth without enterprise contracts.

The step-by-step Clay scoring workflow

Here's the practical sequence for building account scoring inside Clay.

1. Import and deduplicate

Bring in accounts via CSV upload, Clay's built-in company search, or a webhook from HubSpot or Salesforce. Run deduplication before any enrichment columns fire. Running enrichment on duplicates wastes credits, and credits disappear fast if you're not watching.

2. Firmographic enrichment waterfall

Build enrichment columns using waterfall logic. In each column's settings, add the condition "Only run if [column] is empty," then stack fallback providers. Sequence cheapest sources first:

• Employee count: Apollo (primary), Clearbit (fallback)
• Annual revenue: Crunchbase (primary), Apollo (fallback)
• Funding stage: Crunchbase
• Industry and HQ location: Apollo (primary), Clearbit (fallback)

3. Technographic enrichment waterfall

For tech stack signals, use BuiltWith as the primary source. For competitor tools installed, use HG Insights as primary with BuiltWith as fallback.

HG Insights describes intensity as repeated detections of a product signal over time at a company or location, which helps indicate technology usage trends. That's a usable scoring input on its own.

4. Signal enrichment

Layer in time-sensitive signals beyond static firmographics:

• Job postings via Owler or a Google Jobs scraper. A company posting a "Marketplace Operations Manager" is signaling investment in partnerships.
• Website keywords via Clay's web scraper for qualitative ICP signals like "remote-first" or "enterprise security."
• Recent funding via Crunchbase, filtered to funding within 90 days.

5. Build your baseline scoring formula

Add a formula column (type: Number) using Clay's JavaScript-like ternary syntax. Here's a real employee count scoring formula from Clay's documentation:

{{employee_count}} > 300 ? 5 : {{employee_count}} > 100 ? 3 : {{employee_count}} > 50 ? 1 : 0

Key syntax rules: use {{column_token}} to reference columns, && for AND conditions, .includes("text") for keyword search, and always end with : 0 for unmatched cases. Check conditions largest-to-smallest when using > operators.

If you're not comfortable writing formulas manually, Clay's AI Formula Generator translates plain-English ICP criteria into working code.

6. Gate expensive enrichment behind baseline scores

Once a baseline firmographic score exists, add the condition "Only run if [baseline_score] >= X" to expensive enrichments like Bombora intent data, HG Insights depth scoring, and Claygent AI research columns. This prevents burning credits on accounts that clearly don't fit.

7. Run Claygent for qualitative signals

For signals that formulas cannot capture (implied business model, competitive context, qualitative website language), add a Claygent column with a scoring prompt. Set the output format to JSON with score and reasoning fields so individual components feed into downstream formula columns.

8. Build the composite score and tier labels

Sum all sub-scores in a second formula column:

{{employee_score}} + {{funding_score}} + {{tech_stack_score}} + {{job_posting_score}}

Then add a tier label column (type: Text):

{{total_score}} >= 40 ? "Tier 1" : {{total_score}} >= 20 ? "Tier 2" : "Tier 3"

9. Route to CRM and sequences

CRM sync sends scores back to Salesforce or HubSpot automatically on the Growth plan and above. Above-threshold accounts route to AE calendars or outbound sequences. Below-threshold accounts enter automated nurture or get deprioritized until new signals fire. A prioritized list that still requires manual review captures only a fraction of the value.

For teams managing multiple scoring tables, Clay Functions let you build scoring logic once and reuse it. When ICP criteria change, updating the Function propagates across every table calling it.

The signals that matter for $20K+ ACV deals

For mid-market and enterprise SaaS with longer sales cycles, a three-score system is worth copying: fit score, engagement score, and contract value score.

Weight your composite score across four signal categories:

• Firmographic fit (25 to 30%): Industry vertical, employee count, revenue band, funding stage, headcount growth rate. These determine whether an account can structurally support a $20K+ deal before any intent data enters the picture.
• Technographic fit (~20%): CRM type, complementary tools, competitor tech installed, overall stack maturity. Some practitioner teams use website-technology signals to identify prospects that standard prospecting workflows may not reveal.
• Intent signals (~15%): Third-party keyword research surges, competitor review activity on G2, pricing page visits, demo requests. Combining first-party engagement with third-party intent identifies accounts showing real purchase readiness. Either signal in isolation can misread buying readiness.
• Buying triggers (~10%): New funding round, new CRO or CMO hire, headcount growth spike, relevant job postings. These are time-sensitive signals that a buying window has opened.

Negative signals matter just as much. Accounts in excluded industries get hard-disqualified. Company size outside your addressable range scores to zero. Competitor employees get excluded from outbound. Existing customers route to CS. Single low-level contacts get flagged for warming before outreach.

A 90-day window also changes how engagement should score. An account with three employees on the free tier who attended a Clay event last month scores differently from one that checked the website 18 months ago.

Five mistakes that waste your scoring investment

Watch for these patterns. Each one quietly degrades scoring accuracy until reps stop trusting the output.

1. Scoring in isolation from sales: This is the most common way scoring projects die. Marketing scores leads against one set of criteria while sales qualifies against different standards. Both teams can report accurate numbers against their own definitions while the handoff breaks down. Get sales involved in model design before deployment.
2. Single-source enrichment: One provider alone leaves large gaps. A properly configured Clay waterfall setup improves coverage materially. The coverage gap becomes the scoring gap.
3. Over-indexing on firmographics: Models that weight job title and company size too heavily create systematic bias toward structural fit while ignoring behavioral signals. Balance firmographic fit with engagement and intent.
4. Ignoring credit economics: Teams running multi-provider waterfall enrichment without monitoring can exhaust plan credits fast. Structure waterfall order by cost-per-match. Gate expensive enrichments behind baseline score thresholds.
5. Treating scoring as a one-time project: Static models rot. Your ICP evolves, your product positioning shifts, and the scores stop meaning anything. Calibrate quarterly against closed-won data.

When Clay is the right tool (and when it isn't)

Stage of growth determines whether Clay is the right scoring engine for your team.

Pre-PMF or seed-stage companies should stick with CRM-native rule-based scoring. You're still figuring out who buys. A scoring tool won't fix that. Series A through B companies ($2M to $15M ARR) benefit most from Clay as a primary scoring and enrichment engine layered on top of CRM-native scoring.

At Series C and beyond, Clay and enterprise intent platforms like 6sense or Demandbase become complementary rather than substitutes. You can run 6sense or Demandbase for proprietary intent data, with Clay for enrichment, custom scoring, and outbound execution.

Be honest about platform limitations. 6sense's scoring methodology is only partially transparent to users and can feel like a black box. If you route exclusively on their intent scores, you're introducing unverifiable assumptions into your pipeline. Layer their output with first-party signals and firmographic fit rather than treating any single platform's score as ground truth.

Build your Clay scoring engine with Understory

Clay account scoring only generates pipeline when it's connected to coordinated execution across paid media, outbound, and creative.

We're a Clay Enterprise Partner, and we build signal-based outbound systems and ICP scoring workflows for B2B SaaS organizations every day. We handle the Clay buildout, the CRM sync, the outbound sequences, plus LinkedIn and Google ads, coordinated as a single allbound motion.

Book a consultation to scope a Clay account scoring buildout for your ICP.

Frequently asked questions

How long does it take to build an account scoring model in Clay?

A functional baseline model with firmographic and technographic enrichment, formula columns, and tier labels typically takes one to two weeks for a defined ICP. Adding Claygent qualitative scoring, intent signal layering, and CRM sync usually extends the buildout to three to four weeks. The longest variable is ICP clarity, not Clay setup time. Teams without a documented ICP spend most of the project relitigating fit criteria rather than building columns.

How much do Clay credits cost for a typical account scoring buildout?

Credit burn depends on enrichment depth and account volume. A waterfall with four to five firmographic providers running against 10,000 accounts can consume 30,000 to 50,000 credits before scoring formulas even fire. Gating expensive providers (Bombora, HG Insights depth scoring, Claygent) behind a baseline firmographic score is the single biggest credit saver. Most Series A through B teams land on the Pro or Explorer plan and stay credit-efficient by sequencing cheapest providers first.

Can Clay replace 6sense or Demandbase entirely?

For most Series A through B companies, yes. Clay covers firmographic enrichment, technographic signals, custom scoring logic, and CRM sync at a fraction of the cost. Where 6sense and Demandbase still win is proprietary third-party intent data (anonymous buyer research signals across their data co-ops). At Series C and beyond with enterprise budgets, running both in parallel is often more cost-effective than picking one: intent platforms for signal, Clay for execution.

How often should account scoring models be recalibrated?

Quarterly at minimum, with a closed-won review every month. Pull your last 90 days of closed-won deals, compare their scores at the time of routing against their actual conversion outcomes, and adjust weights where the model under-predicted or over-predicted. ICP drift is the most common cause of scoring decay: a model built when you sold to 50-person startups will score poorly once you move upmarket to 500-person mid-market accounts. Sculptor automates this calibration if you have enough closed-won volume to train on.