Apply Checkpoint

The apply checkpoint step verifies details identified in a previous step according to validation formulas, and triggers a review for failed validations.

Rules are specified in validations and connected as modules to the apply checkpoint step in a flow.

When validation fails and a review is triggered, the flow job is paused at the checkpoint until a reviewer confirms or corrects data in Flow Review and resumes the flow. You can modify this behavior by enabling straight-through processing.

As a best practice, insert a checkpoint at these points in a flow:

• After classification but before branching, to ensure that records are routed to the correct branch for data extraction.

• After extraction, refinement, and/or redaction, to ensure that extracted data is accurate.

• Anywhere that data accuracy is critical. For example, add a checkpoint before inserting values into a database.

Straight-through processing

By default, the apply checkpoint step pauses all documents in the batch when any one document fails validation. This behavior ensures that all steps after the checkpoint are executed with the complete batch of documents, but it can slow processing overall.

To change this behavior, in the apply checkpoint step, set Enable Straight-Through Processing to Yes.

With straight-through processing enabled, documents that pass validation continue executing regardless of validation failures in other documents in a batch. When the validation failures are corrected and the flow is retried, the flow runs again from the checkpoint step with all documents, including any documents that originally passed validation. This behavior ensures that any reduce operations are executed with the full batch of documents, but it means that some steps might be re-executed.

If you choose to enable straight-through processing, you can account for the potential re-execution of steps in these ways:

• Make sure any custom UDFs are built to correctly handle repeat execution.

• In Flow Review, use the Finalized Records filter to review only documents that aren’t subject to re-execution of steps. Documents subject to re-execution display a warning indicating that the record can’t be corrected because it will be regenerated on the next flow resume.