Promo code: payback with self-selection
Splits orders with a code into “would have bought anyway” and truly additional ones, calculates the code’s profit, the discount lost on “own” buyers and the threshold share of “would have bought anyway” above which the code loses money.
At a glance
- What question does it answer?
- Does a code pay back if part of the buyers would have bought without it.
- What data is needed?
- Number of orders with the code and the estimated share who would have bought anyway (from a control).
- How to read the result?
- 1,200 orders with the code, basket 6,000, discount 15%, margin 55%. At 45% “own” buyers there are 660 additional orders with a margin of 1,584,000, a loss on “own” buyers of 486,000 (of 1,080,000 discount given) and costs of 30,000: the code earns 1,068,000. The threshold share of “own” buyers is 72%: at 80% the code already loses money (−318,000). Example on the default values: Profit of the code — 1.07 m RUB; Additional orders — 660; Discount given — 1.08 m RUB; Of which lost on “own” buyers — RUB 486,000.
Why and how to reason
A promo code gives a discount to everyone who enters it, but benefits only those who would not have bought without it. So orders with the code split into two groups: would have come anyway (45% of 1,200 = 540) and truly new (660). New ones give margin at the discount: ticket RUB 6,000, margin 55%, discount 15%, so revenue 5,100 and margin 5,100 − 2,700 = RUB 2,400 per order, 660 × 2,400 = 1.58 million. The “own” ones only lose the discount: 540 × 6,000 × 15% = 486 thousand ₽. Minus distribution costs 30 thousand: code profit RUB 1.07 million. Of the 1.08 million discount issued, 486 thousand is lost. Threshold share of would-have-bought-anyway: the code pays while it is below 72%. Profit per rouble of discount 0.99.
When to use it
It is useful before launching a code and after it is issued in mailings, partner sites and coupon sites. Estimate the would-have-bought share from a control group or from the share of existing customers among code users. It does not fit a cart-recovery code (almost all are “own” by definition) nor a sign-up code where there is no “anyway”. A code found by coupon sites has the highest share of “own” buyers.
How to read the result
The main number is the threshold share of would-have-bought-anyway (72%) against your estimate (45%). A margin of 27 points means the code is robust to estimate error. If the threshold is close, the code is fragile: check sensitivity at 60% and 80%. “Lost on own” shows the price of giving the code away without limits. Decision: restrict the code by audience (only inactive, only new), by minimum basket, by period, instead of handing it to everyone. After launch track the share of code use among existing customers: if it rises fast, the code has spread to own buyers and should be restricted.
Common mistakes
Counting all code orders as additional; not knowing the anyway share and entering zero; forgetting distribution costs and partner commission; ignoring that the code leaks to coupon sites and reaches own buyers. Re-check the anyway estimate with a test. Do not count a code activation as an additional order without checking.
What data is needed?
- Number of orders with the code and the estimated share who would have bought anyway (from a control).
- Average basket before discount excl. VAT, the code discount and order margin at full price.
- Optional: distribution costs and the number who received the code.
Inputs
| Field | Unit | Value in the example | Meaning |
|---|---|---|---|
| Orders with the code | orders | 1,200 | How many orders were placed with the code over the period. |
| Share who would have bought anyway | % | 45 | An estimate from a control group without the code or from customer history. Without a control take a cautious, high estimate: self-selection is almost always there. |
| Average basket before discount, excl. VAT | ₽ | 6,000 | Price excl. VAT: the general rate is 22% from 01.01.2026. The rate follows the shipment date; confirm your regime and rate with an accountant. |
| Code discount | % | 15 | The discount of the code as a percent of the order amount. |
| Order margin at full price | % | 55 | Share of revenue excl. VAT left after cost. |
| Cost of distributing the code (optional) | ₽ | 30,000 | Mailing, partner commission, advertising. Empty: zero. |
| Received the code (optional) | people | 20,000 | For the usage rate. Empty: not calculated. |
How it is calculated
| Step | Formula |
|---|---|
| Additional orders | orders with the code × (1 − share who would have bought anyway) |
| Loss on “own” buyers | orders × share × basket × discount |
| Profit of the code | additional × margin with discount − loss − costs |
| Threshold | share = (orders × margin with discount − costs) ÷ (orders × full-price margin) |
Worked numbers
Default values from the app; in the interactive version you enter your own.
- Profit of the code
- 1.07 m ₽Margin of additional orders − discount given to “own” buyers − costs.
- Additional orders
- 660
- Discount given
- 1.08 m ₽
- Of which lost on “own” buyers
- 486,000 ₽
- Profit per unit of discount
- 0.99 ₽
- Threshold share “would have bought anyway”
- 72 %Above this share the code loses money.
- Code usage rate
- 6 %
Profit at another share of “own” buyers
| Share “would have bought anyway” | Code profit |
|---|---|
| 20 % | 2.06 m ₽ |
| 40 % | 1.27 m ₽ |
| 60 % | 474,000 ₽ |
| 80 % | −318,000 ₽ |
A code that reaches coupon sites is handed to people who would have bought anyway: the share of “own” buyers grows. Watch stacking of a code with a shelf discount: see “Stacking”. State the code conditions in full in advertising (Art. 5 of Law 38-FZ); have a lawyer check the text.
Nuances and limits
- The share who would have bought anyway is an estimate: without a control group it is understated.
- Coupon sites hand the code to people who would have bought anyway.
- Do not allow a code to stack with a shelf discount without a margin check.