Use the Job's CSV results to distinguish file processing from successful Salesforce writes. A Completed import can still contain ordinary row errors.

Open the Job and select its CSV Import Action. Review the pinned Data File, processed source rows, file progress and recorded batches. Select each Data Target Action separately to inspect its succeeded and failed record counts.

Source-row totals and Target record totals describe different things. One source row can produce several records through nested arrays or several Targets. A Target with chunk-level work can also contribute records without a one-to-one source-row outcome.

Progress tracks the committed byte checkpoint through the source file. It is not a record success percentage.

The CSV Import Action shows file progress; the Target Action shows Salesforce record outcomes.

The results view displays 20 batch cards at a time, newest first. Use the pagination controls to inspect older chunks. CSV Import cards show the source-row range, status and batch runtime. Target cards show that Action's record outcomes and duration. Each provides a Download batch results link.

Version 1.53 provides per-chunk downloads. There is no combined browser download for the entire import.

Each results CSV retains the original input values and adds:

If an input header conflicts with a reserved name, the generated prefix gains another underscore. Results use comma-separated UTF-8 CSV even when the source used another supported delimiter.

Download the batch results to inspect each source row, its created record ID or its error.

Succeeded means the correlated Target outcomes succeeded. Failed can include a mixture of successful and failed outcomes: inspect the outcomes column before rerunning the row. A successful record created by another Target may already exist.

Processed means the row was processed without a correlated Target outcome. It is not evidence that a Salesforce record was written.

Rolled Back identifies source rows from a chunk whose transaction did not commit. Earlier chunks can still have committed successfully. Recovery reporting may be incomplete if the source cannot be read again or Salesforce cannot store the result file; check the Job error.

In the verified three-row Batchable example, the parent Job completed with two successful Account writes and one missing-Name error. Batch 2 succeeded after the row error in Batch 1. See the worked recovery example for how to prepare a correction without repeating successful Insert rows.

CSV Action metrics aggregate measured work across chunks. The view shows per-Action counts and averages; sampled heap also records the largest observed sample. Queue waiting time is separate from Action duration.

The parent Job's general metric fields describe the latest transaction. Do not treat them as whole-file totals.

Historical CSV Jobs can contain only their final chunk snapshot. The UI identifies these Jobs and does not manufacture missing aggregate results or downloads.

Deleting a Job triggers cleanup of its CSV chunk records, including queued cleanup for larger sets. The result files remain ordinary Salesforce Files and need a separate retention policy.

Review outcomes before preparing a correction file. See Troubleshoot and recover CSV imports.