Data transformation

What the data looked like before, and after processing

The source export gave us one paragraph of free text per rejection and a coarse CRM reason code — impossible to aggregate. Each remark was read and converted into structured columns: a root cause, a sub-reason, an outcome, the income and ticket size mentioned, credit flags and a one-line summary. Same 18,209 rows, now fully queryable.