Correcting and reloading safely

RELOADING A CORRECTED FILE OVER A PARTIAL LOAD IS WHERE DUPLICATES COME FROM. Eight thousand of ten thousand rows imported, you fix the two thousand, and you reload the whole file — and now eight thousand records exist twice. So the question before any reload is: what already landed? Answer it from the APPLICATION, not from the error report. The error report tells you what was rejected; it does not tell you what succeeded, and the two are not complements when a run was interrupted rather than completed. The safe sequence. ONE: count what exists in the application for the batch you loaded — by a key you control, such as a batch reference or a distinctive source identifier. TWO: reload ONLY the rows that are absent, which means filtering the corrected file rather than resubmitting it whole. THREE: count again and confirm the total is what the source says,…

In this module: Module 7 · FBDI: the file-based loader end to end

1. 1The FBDI sequence, and why order mattersFree preview
2. 2Finding and reading the right template
3. 3Mandatory columns, and the ones that only look optional
4. 4Preparing the file without corrupting it
5. 5Loading the interface tables
6. 6Running the import and reading the output
7. 7Errors: finding the row that failed
8. 8Correcting and reloading safely
9. 9Master data versus transactional loads
10. 10Volume: loading a hundred thousand rows
11. 11Proving the load landed
12. 12Rollback, and why there usually is not one
13. 13What breaks: six FBDI failures
14. 14Lab briefing · A supplier and invoice load