Skill market 3 skills

Two periods of operating data plus this period's action list go in; five things come out: metric-change attribution with hypotheses still to be validated, threshold alerts with a troubleshooting order, a one-page three-part review, three experiments with pass / fail / negative decision lines, and reusable lessons.

Map ad reports exported from Amazon Ads / Google Ads / Meta onto one shared column set, split the funnel into four stages (impressions, clicks, conversions, break-even), work out which stage leaks money, how much it leaks and what your bid ceiling is, and output a negative keyword list you can add as-is and a prioritized action table

Run one returns report through three outputs: six root-cause classes grouped by what you can actually fix (size / not as described / damage / expectation gap / logistics / other), the 8 costs behind every single return with an addition formula you can do by hand, and field-level edits telling you which part of the listing to change