At a 2% conversion rate, moving to 2.1% means 5% more orders. That's roughly €500,000 in revenue with traffic and order value unchanged. The lower your starting rate, the more a tenth of a point is worth.
The benchmark
The formula fits on one line: gain = revenue × (rate increase ÷ starting rate).
| Starting rate | Gain for 0.1 point | On €10M |
|---|---|---|
| 1.5% | 6.7% | €667,000 |
| 2.0% | 5.0% | €500,000 |
| 3.0% | 3.3% | €333,000 |
Use this table to decide. To keep score, use revenue per visit, which captures order value as well as rate.
At a footwear brand the authors work with, a test of the label on the cart button moved revenue per visit from €0.925 to €0.944.
Two cents, multiplied by the control group's 14,270 visits over thirteen days, comes to €260, or about €7,000 a year. The testing tool reported double that, and that's how it was first read; the lower of the two figures is the one to keep.
The rule
Translate this gain into margin before you talk about it.
These extra orders come with no additional ad spend, since the traffic is already paid for. Their margin is therefore better than that of your average order, on one condition: that your returns are deducted.
Check that your figure excludes them: by default, both Shopify and GA4 count the order, not the return.
Leadership, for its part, already thinks in terms of margin. One executive put it this way in September, about a free-shipping threshold to test:
"It costs me 5, it brings me 10, so I gain +5. You need to be able to compare yourself to something, because otherwise giving things away will always look better than not giving them away."
The steps
- The base. Take your online revenue and your conversion rate measured on human sessions.
- The formula. Apply it for 0.1 point.
- The margin. Multiply by your contribution margin rate, returns deducted, excluding acquisition.
- The discount. Count only part of the measured gain: the lift observed on a winning test is almost always overestimated, and it wears off over time.
- The comparison. Set the result against the cost of the project.
A sentence to say to your leadership, to reuse with your own numbers:
"A tenth of a point is €500,000 in revenue and, for us, about €150,000 in margin. We'll count only half of it until a test has confirmed it."
The trap
A tenth of a point is also the normal week-to-week variation. Seeing the rate go from 2.0% to 2.1% on a Monday proves nothing.
This calculation is for deciding whether a project is worth launching, never for announcing a win. To call a win, you need a test or several weeks of hindsight, and a discount.
For a premium men's fashion brand with under a million euros in online revenue, a program of ten tests over six months was costed out with these rules: one winning test in five based on past results, one in three in good years; an effect counted at +10% whatever gain the tool displays; a 20% discount in the first year; and an effect that fades out over twelve months.
With these rules, the projection came to between four and six times the cost of the program; without them, it showed a figure nobody would have believed.
