Understand the source before changing it
A spreadsheet may contain hidden conventions that the column headings do not explain. A blank cell might mean “not checked,” while a dash means “not applicable.” A second row might be a new item or a correction to the first. These differences need an agreed interpretation.
We begin with an inventory of the available sources, their owners, and their intended use. Early review can use a redacted sample rather than a complete operational dataset.
Make mapping decisions explicit
Mapping connects each source field to a destination field or a documented exclusion. Changes to date formats, names, status labels, or categories should be described in plain language. A transformation that seems obvious to one person can alter the meaning for another.
Some values will not fit a reliable rule. These should appear on an exception list for human review. Automatically replacing uncertainty with a confident-looking value is not a successful cleanup.
Review duplicates without losing context
Similar names are a useful signal but do not prove that two rows represent the same item. A duplicate review may consider stable references, source history, and supporting context. The person deciding a merge should understand which details would be retained.
The agreed method should also cover what happens to links that point at a retired record. A visible relationship to the surviving entry can be more useful than silently removing the old identifier.
Reconcile the result
A small test import should be reviewed before a broader move. Compare source and destination counts, inspect a representative set of records, and account for exclusions and exceptions. Counts help reveal omissions, but they do not prove that every field is correct.
The delivery can include a mapping document and a reconciliation summary. Access permissions, retention decisions, and recovery procedures need agreement with the information owner. Ongoing imports or synchronization require their own defined operating arrangements.