Diagnostics, sensitivity tests, and worked examples on ecommerce CRO, paid media, P&L strategy, returns and discount discipline, A/B testing, and driver-based forecasting. Written for the operators in the weekly review, not the people building decks for it.
When CVR plateaus while sessions climb, the reflexive answer is "traffic quality." It's almost never that. Here's the math behind the plateau and a six-step diagnostic that finds the real cause.
Reported ROAS holds steady while CAC keeps climbing. The metric isn't broken — it's incomplete. The contribution-margin lens that tells you whether paid media is buying growth or just buying orders.
P&L data without a framework is just noise. The contribution-margin tree, the four levers, and a sensitivity ranking that turns three competing roadmaps into a single ranked agenda.
Together, they erode 4–8 points of contribution margin a year while every dashboard says revenue is up. The gross-to-net waterfall, the discount-stacking breakdown, and the cohort math that catches it.
A roadmap of thirty tests with no strategy is a tax on engineering and a placebo for the team. How to design a program with hypotheses, prioritization math, statistical power, and actual learning velocity.
"Last year + 20%" is a wish, not a forecast. How to build a driver-based model — sessions × CVR × AOV with elasticity assumptions — that produces a three-scenario forecast you can rerun in ten minutes.