Sort an example store's 137.947 orders by how deeply they were discounted and something counterintuitive appears: the discounted orders are worth more. Undiscounted orders average €93 of net revenue, orders discounted up to ten percent average €102, and even orders discounted by more than twenty percent average €95. Read carelessly that is an advertisement for discounting. Read properly it is a lesson about what a correlation in your own order book can and cannot tell you, and the honest conclusion points in a rather different direction.
- Undiscounted orders: 89.972 of them, averaging €93 of net revenue.
- Discounted up to 10%: 16.015 orders at an average depth of 9,7%, averaging €102.
- Discounted 10% to 20%: 23.519 orders at 16,8% depth, averaging €98.
- Discounted over 20%: 8.441 orders at 25,4% depth, averaging €95, so the advantage narrows as depth grows.
What the numbers say
Two thirds of orders carry no discount at all, and they are the smallest by net value. Every discounted band is larger, and the advantage shrinks as the discount deepens: nine euros above the undiscounted average at shallow depths, five euros above it beyond twenty percent.
So the shape is real and the causal story is not established. Threshold-based promotions, bundles and codes that apply to larger carts all select for bigger baskets, which means the discount may be following the basket rather than creating it.
| Depth | Orders | Average depth | Net revenue per order |
|---|---|---|---|
| No discount | 89.972 | 0% | €93 |
| Up to 10% | 16.015 | 9,7% | €102 |
| 10% to 20% | 23.519 | 16,8% | €98 |
| Over 20% | 8.441 | 25,4% | €95 |
Why bigger orders do not mean better orders
Net revenue per order is not contribution per order. A discounted €102 order gave away part of its margin to get there, and the leverage is brutal: because contribution is a thin slice of the price, a ten percent discount typically removes about a quarter of what the order keeps.
Applied to these bands, an order that is nine euros larger and ten percent cheaper is not obviously ahead, and one that is two euros larger and twenty-five percent cheaper is clearly behind. That is the arithmetic a discount impact calculator does properly, on your own margins rather than on an average.
The selection problem, stated plainly
Any comparison between discounted and undiscounted orders in your own data is a comparison between two groups you did not assign. Customers who use codes differ from those who do not, promotions run at different times of year, and threshold offers explicitly require a larger basket to qualify.
So the correct reading of this study is a warning rather than a finding: a store that looks at the same table and concludes that discounting raises basket size has skipped the step where somebody asks whether those customers were already going to spend more.
What would actually settle it
A comparison across time rather than across groups: total contribution in a promotional period against a comparable period without one, adjusted for whatever else moved. Not revenue, and not average order value, both of which a promotion reliably increases while contribution falls.
That is a modest experiment and almost nobody runs it, which is why so many promotional calendars survive for years on the strength of a table like the one above. The framework for judging one properly is in discount strategy, which starts from the volume a code owes you before it runs.
What the depth bands hint at
One pattern in the table is worth taking seriously even under all the caveats above: the advantage falls as depth rises. Shallow discounts sit on the largest baskets, and the deepest band has both the fewest orders and the smallest advantage over undiscounted ones.
That is consistent with a story most merchants will recognise. Shallow codes ride along with carts that were already large, while deep codes are doing clearance work on smaller, more price-driven orders. If that is what is happening in your own data, the two kinds of promotion deserve to be judged separately rather than averaged into one discount rate.
The one honest conclusion
The data supports a narrow claim: on this store, discounts concentrate on larger baskets, and the deeper the discount the less that advantage holds. It does not support the claim people want, which is that deeper discounting produces bigger orders. Distinguishing those two is most of what separates a useful analysis from a comforting one, and it is why every study on this site says what its numbers cannot show as well as what they can.
The measure that decides any of this is discount depth against contribution rather than against revenue, computed on your own economics; how every figure is computed sets out the rules behind these ones.