Treat a batch as a set you can reconcile
Batch invoice processing lets you submit several invoices under one extraction schema. The practical benefit is consistent columns across the set. The review problem is making sure every source file is accounted for, including documents that fail and retries that might otherwise become duplicates.
Parzable accepts multiple files in the Extract view and processes the queue sequentially. You can inspect individual results and download completed results as a batch CSV. Do not interpret an active queue as parallel processing or assume an unlimited batch size.
Start with a batch you can review promptly. Check your available page allowance first: a multi-page PDF can consume more than one page. Keep unrelated invoices in separate files and keep each invoice’s continuation pages together.
Use a manifest outside the extraction output
A manifest is a simple checklist of expected files. Keep it separate from the CSV so a failed extraction cannot disappear from your intake record. The blank batch manifest provides columns for batch name, filename, expected pages, extraction status, export status, review status, and notes.
- Record the source filenames and expected page counts.
- Choose one schema for the batch. Separate files needing different columns.
- Upload the files and keep the processing view open while the queue runs.
- Record completed files and errors in the manifest.
- Download completed results, then compare filenames with the manifest.
- Review the values before marking a record ready for its destination.
The manifest is a downloadable working template, not an automatic Parzable reconciliation feature. Fill it using the statuses you observe.
A worked reconciliation
Fictional example: you receive 12 invoice files. Ten complete and two fail. The completed-result CSV should contain ten data rows if there is one result per file. If you expected 12 rows, resolve the two failures rather than treating the export as complete.
| Checkpoint | Example count | Question |
|---|---|---|
| Files received | 12 | Does the manifest list each file? |
| Completed / failed | 10 / 2 | Do these account for all 12? |
| Exported data rows | 10 | Are these the ten completed files? |
| Reviewed / held | 9 / 1 | Is the held result excluded from the final import? |
After retrying the two failed documents, merge their reviewed results once. Do not blindly append another download of the original ten records.
Resolve exceptions without hiding them
For an error, record its message and check the file, schema, account allowance, and page completeness before retrying. For a wrong extracted value, retain the original and document the correction in your review sheet. For repeated failures on one supplier layout, isolate a small sample and improve the field descriptions before returning to the full batch.
Use a clear batch name, such as a client identifier and month, without exposing unnecessary personal data in filenames. Keep a link between source, extracted record, and reviewed destination record. Extraction completion, review completion, and import completion are separate checkpoints; your batch is finished when the intended destination and all unresolved exceptions are accounted for.
Put the workflow to work
Try your own invoice in Parzable
Create an account, define a schema, and test a small sample. Review extracted values before using them. See current plans and page allowances.
Try Parzable