# Clay workflow acceptance example
Growth Cab | Version 1 | 10 September 2026

This is a simulated acceptance fixture, not a client result, Clay template export or a live enrichment benchmark. All company domains end in .example and must not be contacted. No real contact records or provider calls are included.

## Files and sample policy
Import clay-input.csv into a scratch table. Compare your decision columns with clay-expected-output.csv. Keep outbound integrations disconnected. Sample ICP: US B2B SaaS, 50–500 employees, VP Sales role. These invented criteria illustrate a workflow; they are not Growth Cab's universal ICP.

## Ordered workflow
1. Normalize company domains and role labels. Preserve record_id. Mark duplicate account-role pairs after the first occurrence.
2. Apply suppression before any paid lookup. A suppressed record cannot progress even if it fits the ICP.
3. Reject explicit ICP mismatches. Do not treat missing values as a match or fill them with a model's guess.
4. For eligible or incomplete records, query a primary enrichment source only for fields needed by the decision. Use a fallback only when evidence is missing or contradictory. Retain provider, retrieval date, field value and evidence location.
5. Route unresolved records to hold, with an owner and reason. Accept_for_review means fit has passed this sample policy; it does not mean permission to contact, a verified email, a lead or an opportunity.
6. Human review checks evidence, current role, suppression, contact suitability and downstream requirements. Only a separate approved campaign gate can permit export or sending.

## Expected reconciliation
Six input rows = one accepted for review + three rejected + two held. S02 duplicates S01; S03 fails the employee threshold; S04 is suppressed. S05 lacks size evidence and S06 has conflicting role evidence. None are campaign-ready.

The fixture uses one simulated unit for a primary lookup and two additional simulated units for a fallback. S01 uses 1; S05 and S06 use 3 each. Total 7 units / 1 accepted-for-review record = 7 units per accepted record. These are arithmetic teaching units, not actual Clay credits or vendor prices. If accepted count is zero, the ratio is undefined; report the spend and zero accepted records separately.

## Acceptance checks before expanding
All six IDs reconcile exactly once; suppression and duplicate rows have zero lookups; no unknown value becomes a fabricated fact; no held record exports; every acceptance has inspectable evidence. Replace the fixture with an authorized representative sample before making any performance or cost claim.

Source page: https://www.growthcab.com/clay-consultant