Boards rarely lack ecommerce data. They lack a small set of numbers that everyone agrees means something. When reporting runs to forty slides, the interesting movement is usually in the four that nobody put on the same page.
1. Blended contribution, not channel ROAS
Revenue after cost of goods, fulfilment and total marketing investment — for the whole business, in one line. Channel-level return can improve while this falls, which is the single most common way a growth problem stays invisible for two quarters.
2. Cost to acquire a new customer
Total marketing investment divided by genuinely new customers. Not blended CAC across a repeat-heavy base, which flatters efficiency whenever retention is doing the work. If this rises while contribution is flat, you are buying volume at the expense of margin.
3. Repeat rate within a fixed window
Share of customers who buy again inside a window that matches your real purchase cycle — 90 days for consumables, twelve to eighteen months for considered categories. Lifetime value models hide too many assumptions to argue with; a fixed-window repeat rate does not.
4. Non-branded demand
Branded search and direct traffic measure the demand you already created. Non-branded discovery — organic, assistant surfaces, marketplaces, comparison — measures whether you are still reaching people who do not yet know you. When it flattens, growth becomes a harvesting exercise with a ceiling.
Every stalled ecommerce business we look at was reporting healthy numbers. They were the wrong four.
How to read them together
- Contribution flat, new customer cost rising: an acquisition economics problem, not a creative one.
- Contribution rising, non-branded demand flat: you are harvesting; the ceiling arrives in two to four quarters.
- New customer cost stable, repeat rate falling: a product, merchandising or lifecycle problem being paid for by paid media.
- All four flat while activity increases: a prioritisation problem, and the cheapest one to fix.
None of this requires new tooling. It requires agreeing on four definitions and holding them constant for long enough that a trend means something.