Back to dashboard

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.

Before

10 raw columns, 1 free-text paragraph

No way to group, count or trend

Processing

Every remark read and classified

Taxonomy of 10 root causes + sub-reasons

After

13 structured, typed columns

Filterable, chartable, coachable

Showing the 0 most recent records