Marketplace dashboards often disagree on the same SKU because attribution windows differ. Start by locking a single primary metric—conversion on the PDP—and note supporting signals like add-to-cart rate without treating them as equal.
Run experiments in short, documented bursts: change one listing element, capture a before snapshot, and resist editing mid-week unless a policy violation appears. Keep a plain-language changelog your merchandising teammates can scan.
Finally, debrief with qualitative notes from customer messages. Numbers tell you direction; verbatim questions tell you what to test next.