$ man clay-wiki/scoring-logic

A scoring system converts raw data into a decision. Should you contact this company? Is this person worth spending credits on? Should this lead go to the sales team or the nurture queue? The score answers all of these questions with a single number. But only if it's built correctly. A bad scoring system is worse than no scoring system — it gives false confidence. A good scoring system makes the right decision obvious at a glance.

PATTERN

The default weighting: 5 points per signal. If 80% of your closed-won accounts have 500+ employees, that's a strong signal — 5 points. If 90% are in SaaS or FinTech, that's 5 points. If 60% use Salesforce, that's a weaker signal — 3 points. If 40% are in New York, that's a weak signal — 2 points. The weighting comes from your data, not from theory. Don't sit in a room and decide that industry is worth more than employee count. Look at your closed-won accounts and let the patterns dictate the weights. The close-won data IS the scoring model.

PATTERN

Most teams build scoring models from assumptions. "We think mid-market SaaS companies are our best fit." Maybe. But do you know? Pull your closed-won accounts. Enrich them. Look at the data. What industries actually closed? What employee counts? What tech stacks? What geographies? The patterns in your closed-won data are more accurate than any theory because they're based on reality — these companies actually bought. Build your scoring model from those patterns. Update it quarterly as new deals close. The model evolves with your business.

PRO TIP

Clay has a scoring integration that produces clean, auditable output. Use it instead of building scoring formulas. Here's why: (1) The scoring integration shows the breakdown — which signals contributed how many points. Sales reps can see exactly why a lead scored 8 instead of 5. (2) Formulas are opaque — a formula that returns "8" doesn't explain itself. The rep has to reverse-engineer the logic. (3) The scoring integration handles edge cases (missing data, partial matches) more gracefully than a formula chain. The score must be self-explanatory at a glance. If someone looks at a score of 8 and can't immediately understand why, the scoring system is broken. The integration solves this by showing the work.

PATTERN

After calculating the total score, bin it into actionable tiers: • 8-10: Tier 1 — High Priority. Route to outreach immediately. These accounts match your closed-won profile closely. • 6-7: Tier 2 — Qualified. Route to outreach with standard priority. Good fit with some missing signals. • 4-5: Tier 3 — Nurture. Don't contact now. Add to nurture sequences or park for re-evaluation next quarter. • Below 4: Tier 4 — Disqualify. Don't contact. Don't enrich further. Save your credits. The thresholds should map to actions. If Tier 2 and Tier 3 get the same treatment, you don't have four tiers — you have three. Every tier should trigger a different downstream behavior.

PATTERN

Every scoring system needs a primary gate — one signal that must be present for qualification regardless of total score. For an Atlassian partner, the primary gate is Atlassian footprint. A company could score 9/10 on every other signal, but if they don't use Atlassian products, they're not a fit. The primary gate prevents false positives. Without it, a company with perfect firmographics but zero product relevance could score highly. The gate catches that. Implement it as a separate check: if primary gate fails, the score is automatically 0 regardless of other signals.

ANTI-PATTERN

Building a scoring model without closed-won data. You're guessing. Look at your actual customers before deciding what signals matter. Using formulas instead of the scoring integration. The formula returns a number with no explanation. The integration returns a number with a full breakdown. Always choose transparency. Making the score too complex. If you have 15 signals each weighted differently with conditional logic, no one will understand the output. Keep it to 5-8 signals maximum. Simplicity scales. Complexity breaks. Never recalibrating. Your ICP shifts. Your product evolves. Your market changes. Re-run closed-won enrichment quarterly and update the scoring model. A stale scoring model actively misdirects your pipeline.