Reconcile missing and mismatched integration records

Define which records belong together

A green transfer indicator does not establish that the target data is correct. Reconciliation needs a defined source set, mappings to target identifiers and fields with comparable business meanings. AWS describes checking corresponding source and target rows. A business integration, however, may create several target documents from one order. Check that relationship rather than just row counts.

Record the selection, version, time zone and observation time. A continuously changing source may not be directly comparable with a target read earlier. SAP explains this boundary for reconciliation. Also check filters, cancellations and units before classifying differences as errors.

Example: one quantity is wrong, one order is waiting

This example is entirely fictional. The unchanged export B-7 is released at 10:00. This workflow has an agreed maximum of 5 minutes until records become visible in the target. Each source order must create exactly one target order; all quantities mean individual units. The target extract is read at 10:03. This is not an atomic snapshot of both systems.

At 10:03, the final order is still within the allowed time. If it is still absent after 10:05, it is overdue and needs investigation. That does not prove its message was lost. A queue, filter or rejection could explain the difference. Equal total counts would not rule out swapped or incorrect contents either.

Make corrections traceable

For each discrepancy, record the identifier, expected and observed values, time context, owner and next check. Collect only the data needed. Before retrying, check whether a target order already exists and how duplicate processing is prevented.

Address the cause specifically, then reconcile the same defined set again. Keep open cases separate from confirmed matches. One successful retry does not validate other orders or later changes.

Sources

• AWS DMS: Data validation

• SAP: Datenabgleich