End-to-end automation of a weekly reporting pipeline

Context

National retail chain (coffee) · weekly + monthly reports

Problem

Every week, thousands of records from 7 channels were cleaned, pivoted and placed into a 40+ slide deck by hand. Consistency between periods depended on the person; when an analyst changed, the standard drifted.

Approach

I built the path from raw CSV to deck as a modular Python pipeline: brand-phrase and noise filters → deduplication → classification → pivot generation → chart/card/text placement into the PowerPoint template → QA audit. Every rule (threshold, exception, font size, bullet count) is tied to a decision record; the pipeline reads values from records, not from memory.

Impact

Weekly preparation dropped from 3–4 hours to 30–45 minutes (excluding data reading, which was previously fully manual and was automated separately); 50+ decision records made the report standard person-independent; the same skeleton was ported to a second and third report family.