A waterfall runs providers in order until one returns a result. Designed well, it maximizes coverage per dollar. Designed badly, it maximizes spend per duplicate row.
Design rules
- Order by expected value: match rate ÷ credit cost, best first. A provider with 85% match at 1 credit beats one with 90% at 5 credits for most use cases.
- Cheap conditional steps are free wins. Select validation providers cost 0 credits under current pricing — validation after finding an email is nearly free insurance.
- Cap the depth. Each additional fallback that adds less than ~5% coverage is usually not worth its call cost. Measure, do not guess.
- Guard expensive steps with conditions. Only run a premium enrichment when a cheaper signal (segment, size, tier) says the record is worth it.
- Record provenance. Which provider produced each value — so quality problems trace to a provider, not to "the data".
Measure the waterfall, not the providers
The number that matters is credits per usable record: total credits spent ÷ records with valid, usable output. A waterfall with worse per-provider match rates but better input normalization often wins on this number.
Try it: the Cost & Usage Calculator models expected calls and credits for waterfall sequences, including conditional fallbacks.
Need help applying this?
Describe your project and the ClayForEnterprises team will review it — the resources stay free either way.
Submit a project Cost & Usage Calculator Waterfall efficiency guide