Clay Actions and Data Credits Explained
Clay Actions measure platform work such as enrichment execution and sending data to other tools. Data Credits pay for data or AI purchased through Clay's marketplace. A workflow can consume both, and using your own provider key can move provider charges outside Clay while platform usage still counts.
By LeadHaste Editorial · AI-assisted editorial review ·
Separate the two costs
Think of an Action as usage of the orchestration layer and Data Credits as marketplace consumption. They answer different budget questions: how much work the platform performs, and what third-party data or AI you purchase through it.
Clay's current documentation distinguishes those categories explicitly. It also describes exceptions for operations such as formulas and certain internal table transfers. Check the current operation and plan rather than assuming every populated cell costs the same amount.
A conceptual lookup
Suppose a row runs a marketplace email lookup and then exports an accepted contact to a CRM. The lookup and export are separate operations. Their usage should be assessed separately; one successful row does not necessarily mean one unit of total usage.
If the lookup uses your own provider connection, the provider may bill your account directly. That can remove the marketplace Data Credit charge for that provider, but does not make the entire process free. Review Clay usage and the provider's bill together when calculating the cost of a usable record.
Why waterfalls and AI need a usage review
A waterfall can attempt several providers before it finds an acceptable result. AI research can vary with the configured model and work performed. A single row can therefore involve more activity than the final visible field suggests.
Use the current estimate and actual run history to understand the operation. Avoid projecting a fixed cost per row from one easy example. Failed or unsuccessful operations can also have operation-specific charging rules, so inspect the documentation for that case rather than generalizing a success-only assumption.
Practical cost controls
Keep existing trusted results, deduplicate records and add conditions before paid work. Use formulas for deterministic cleanup when a provider lookup or AI judgment is unnecessary. Test a bounded sample and compare cost per accepted output, not merely cost per populated field.
Plan limits, rollover and top-up rules can change. Confirm them in the workspace's current billing settings before purchasing or changing capacity. The glossary explains the units; it is not a live price quote or authorization to spend. Store connection ownership and usage assumptions with the workflow so another operator can reconcile where the cost came from.
Sources
Useful next steps
Related terms
Put the pieces into a system you own.
Explore your ICP and campaign fit on a free discovery call, then decide what your outbound operation needs.
Book your free ICP review →