Before this: Map categories with an approved dictionary
The idea
Adding 'web-chat → chat' moves rows from exceptions to a category. Before releasing, count how many rows change and check examples.
Version the mapping table and note who approved the change, so reports can be explained later.
Worked example
Input: new pair: web-chat → chat
- Visual
Map values → add pair → compare before/after counts
- SQL (DuckDB)
SELECT approved, COUNT(*) FROM mapped GROUP BY ALL
- pandas
df_new["channel"].value_counts() - df_old["channel"].value_counts()
Result: chat +14 rows, unmapped −14
Common mistakes
- Unreviewed mapping edits.
- No record of who approved a change.
Check yourself
What should you measure before releasing a mapping change?
How many rows move and to which categories.
Practice it
Map channels with an approved dictionary.