Categories that changed meaning

• A territory, tier, or product line was split, merged, or redrawn, and the old name was reused.
• A threshold moved, so the same label now applies to a different group.
• A code was retired and later assigned to something new.

• Trends show a jump or a drop on the day the definition changed, and it is read as a real change in the business.
• Comparisons between periods set two different groups against each other under one name.
• No check catches it, because every value is a valid label.

• Chart the row count for each category over time. A sudden step with no known cause is a sign.
• Look for categories that first appear partway through the file, and see which older category shrank at the same moment.
• Ask the people who maintain the list whether it has been revised, and when.

• Establish the date of each change from outside the file. The data alone rarely proves it.
• Record which definition each row was written under, in its own column.
• Restate old rows under the new definition only where a finer field, such as state or product, makes that possible.
• Where old rows cannot be restated, say so, and compare periods only within one definition.

• Codes without a lookupA column holds short codes such as 3, A, or XR, and the table that says what each code means is not in the file.
• Inconsistent category labelsA column meant to hold a few fixed values holds many spellings of them.

TermsSchema drift, Reference data, Consistency, Provenance