How do you review a towing call-log CSV without guessing identities?

Establish what the records mean

Start with the reconciliation lab’s synthetic ledger and data dictionary. Its event rows describe contacts; inquiry rows describe requests. A real provider’s export may use different fields for calls, legs, transfers and timestamps, so its mapping must be verified independently. Never upload private records to a public demonstration service merely to inspect them.

A worked example

In the sample, fifty event rows map to thirty inquiry records, with one event intentionally unlinked. Several later events are retries and one is a transfer. Loading the file does not turn fifty events into fifty prospective jobs. The unlinked event appears in the unresolved queue rather than being assigned to a plausible request by phone or time.

Choose the next action

Map the supported columns, resolve structural errors and inspect preserved source rows before reading totals. Compare row counts with the original file. If your provider’s data cannot establish inquiry identity, create a separate human-reviewed mapping instead of inventing a field. Keep the raw export untouched.

Check whether it worked

Export the summary and queue, reopen them and verify a few IDs back to the source. Check that text fields cannot execute spreadsheet formulas and that unknown values remain blank or labeled. Document the export version and mapping so the next review uses the same rules. This workflow is vendor-neutral and does not claim a tested proprietary integration.

A repeatable review order

1. Export source and dictionary. Record the evidence and mark any missing answer as unknown.
2. Event versus inquiry fields. Record the evidence and mark any missing answer as unknown.
3. Column mapping. Record the evidence and mark any missing answer as unknown.
4. Unlinked records. Record the evidence and mark any missing answer as unknown.
5. Reopened export and ID checks. Record the evidence and mark any missing answer as unknown.