Data
Reconcile a data migration
Prove the target matches the source, field by field.
Fill it in
Both sides.
Matching row counts hide truncated columns.
With enough context to diagnose each class.
Your prompt
Reconcile the migration. Compare FIELDS, not counts. Matching row counts is the reassurance that hides everything: a truncated column, a timezone shift, a decimal rounded, a null that became an empty string. All of those preserve the count exactly. Check the aggregates too: sums of every numeric column, min and max of every date. Those catch a systematic shift that a row sample misses. For each class of difference, show enough examples to diagnose it, and say how many rows are affected. Separate expected transformations from unexpected drift, and be explicit that anything you cannot classify is unexplained rather than acceptable.
Continue and pick your folderOpens with everything above already filled in.
Why this works
Matching row counts hide everything that matters: a truncated column, a timezone shift, a null that became an empty string. This compares fields and aggregates, and is explicit that anything it cannot classify is unexplained rather than acceptable.