When a source column changes

Treat a changed field as a question about meaning before changing the mapping.

The file still arrives

An export that previously supplied amount now supplies net_amount. The file reaches its destination on schedule, and a permissive loader can still write the other columns. The next report may show missing amounts or a different total.

A renamed header might refer to the same value. It might also indicate that discounts, tax or another adjustment have been removed. The new name alone does not establish which interpretation is correct.

This is an illustrative source change. Before modifying a real mapping, compare the source documentation and representative old and new records.

Separate structural and semantic checks

A structural check can establish that the required column is present and its values can be read as the expected type. Comparing old and new records is still necessary to establish whether net_amount means the same thing as amount.

A semantic comparison needs records whose treatment distinguishes the possibilities. A discounted transaction is more informative than one with no adjustment. A missing or zero amount may also be valid in some source situations, so a blanket replacement with zero can conceal a gap.

The acceptance rule for the changed feed should identify which conditions are established, which remain uncertain and whether the affected output can be used while that question is resolved.

Follow the field into its consumers

The incoming column may feed a standardized staging field, a transformation and several report measures. Replacing its name in the ingestion code is only part of the change.

Identify the consumers included in the assignment. A measure using gross value may need the earlier field to remain available while another measure intentionally adopts the new value. Reusing one familiar column name for both interpretations makes the distinction difficult to discover later.

Where an old and new source format must coexist, the model should record which interpretation applies. That can prevent a backfill from applying the latest mapping to an older file that used a different definition.

Choose a controlled response

Depending on the feed, an unexpected change may stop the affected load, isolate its output for review or trigger a comparison before the report refreshes. The choice should reflect the consequence of using incomplete data and the systems available to support the check.

Keep the original received material where the agreed retention policy permits it. Record the mapping revision and the examples used to verify its meaning. A future maintainer can then see why the mapping changed, instead of finding only a note that the import was fixed.

Once the interpretation is established, update the checks as well as the mapping. A pipeline that accepts the corrected format should still expose a future change to the fields the reports depend on.