Channel response curve: where advertising stops paying back
From several pairs of “budget and channel revenue” fits a saturating curve, shows how close you are to the ceiling, how much the last unit of money brings and at which budget the marginal return equals the cost.
At a glance
- What question does it answer?
- Where a channel budget stops paying back and whether it should change.
- What data is needed?
- 3–6 pairs of “budget and channel revenue” for periods of the same length.
- How to read the result?
- Over six points (budget 100 thousand to 1m) the curve gives a revenue ceiling of 5.0m with a goodness of fit of 99.98%. At a budget of 600 thousand 73% of the ceiling is reached, the last unit brings 2.94 of revenue, that is 0.32 of margin above the unit of cost. The marginal return equals the cost at a budget of 726 thousand (21% higher): channel profit rises from 1.054m to 1.073m. Example on the default values: Channel revenue ceiling — 5 m RUB; Share of the ceiling reached at the current budget — 73.4%; Revenue per last unit of budget — RUB 2.94; Margin per last unit of budget minus the unit — RUB 0.32.
Why and how to reason
An advertising channel has a ceiling: the first roubles attract the most responsive buyers, and then each new rouble brings less revenue. So channel revenue is described by a saturating curve R(x) = a × (1 − e^(−b·x)): a is the ceiling, b the speed of approaching it. The curve is fitted to pairs of budget and revenue from past periods (6 points, from 100 to 1,000 thousand ₽, fit quality 99.98%). The ceiling comes out at RUB 5.0 million. At the current budget of 600 thousand 73.4% of the ceiling is reached, i.e. revenue of about 3.67 million. The last rouble of budget brings RUB 2.94 of revenue; at a 45% margin that is RUB 1.32 of margin, 0.32 more than the rouble spent: it still pays. Marginal return drops to zero profit at a budget of 726 thousand ₽. Channel profit there is 1.07 million against 1.05 million now: the gain is small but there is no point in paying beyond.
When to use it
Use it for channels with a controllable budget (search, social, marketplace ads) where there is history by period with different budgets. You need at least 3 points, better 6–8, with a visible spread in budget. Assumptions: the channel is stable, season and price do not change the response, no lag. It does not fit a channel working through brand over a long time nor a case where the budget changed together with the campaign: the curve will be confused. It does not replace an incremental test.
How to read the result
See how much of the ceiling is already reached (73%), and revenue from the last rouble: if its margin is above a rouble you can still grow, if below you should cut. The recommended budget (726 thousand ₽) is the point where the last rouble returns exactly a rouble. The profit gain on the flat part of the curve is small, so do not chase accuracy: a ±15% range around the optimum gives almost the same. Decision: raise the budget while marginal return exceeds cost, and move money from saturated channels to unsaturated ones. It is worth checking the curve with a small budget change (plus 10% for 2 weeks) and compare expected and real revenue: if the gap is large, the model does not fit and the decision must follow a test.
Common mistakes
Typical trap: fitting on two or three points; taking periods with different margin or price; treating the ceiling as a real market limit (it is only the response limit of the channel); changing the budget by the full jump at once. A past curve does not guarantee the future: check step by step. Do not enter budgets and revenue for periods of different length.
What data is needed?
- 3–6 pairs of “budget and channel revenue” for periods of the same length.
- The margin on revenue before advertising and the current budget.
Inputs
| Field | Unit | Value in the example | Meaning |
|---|---|---|---|
| Margin on channel revenue before advertising | % | 45 | The share of revenue excl. VAT after cost, returns and commissions. Do not subtract advertising. |
| Current budget | ₽ | 600,000 | The channel budget for a period, as in the table. |
| Threshold: share of the response ceiling at which to warn | % | — | If more is reached, a further budget increase gives little. Empty: no threshold, the color does not change. There is no single standard in Russia: enter your own. |
How it is calculated
| Step | Formula |
|---|---|
| Curve | revenue = a × (1 − e^(−b × budget)); a and b are chosen by least squares |
| Marginal return | a × b × e^(−b × budget) |
| Optimum | budget = ln(a × b × margin) ÷ b: there the margin on the last unit equals the unit of cost |
| Profit | margin × revenue − budget |
Worked numbers
Default values from the app; in the interactive version you enter your own.
- Channel revenue ceiling
- 5 m ₽Revenue as the budget tends to infinity: the curve’s parameter a.
- Share of the ceiling reached at the current budget
- 73.4 %
- Revenue per last unit of budget
- 2.94 ₽
- Margin per last unit of budget minus the unit
- 0.32 ₽Positive: the budget can still grow; negative: the last unit does not pay back.
- Budget where the marginal return equals the cost
- 726,387 ₽ · up 21 %
- Channel profit at the current budget
- 1.05 m ₽Margin on revenue minus the budget.
- Profit at the suggested budget
- 1.07 m ₽
- Goodness of fit R²
- 100 %
Return at different budgets
| Budget | Revenue | Revenue per unit | Margin on the last unit | Channel profit |
|---|---|---|---|---|
| 300,000 ₽ | 2.42 m ₽ | 8.08 | 2.56 | 791,058 ₽ |
| 450,000 ₽ | 3.15 m ₽ | 7 | 1.84 | 968,467 ₽ |
| 600,000 ₽ | 3.67 m ₽ | 6.12 | 1.32 | 1.05 m ₽ |
| 750,000 ₽ | 4.05 m ₽ | 5.4 | 0.95 | 1.07 m ₽ |
| 900,000 ₽ | 4.32 m ₽ | 4.8 | 0.68 | 1.04 m ₽ |
A high R² (100 %) with 6 points proves almost nothing: two fitted constants always fit a few points well. The shape of the curve is confirmed by new periods at a different budget, not by R².
Channel revenue depends on how it is attributed: with last-click attribution the return is overstated. Parameters are fitted only to your data, no external values are used here.
Nuances and limits
- Three points is the minimum for two parameters: the recommendation is very sensitive to noise.
- A suggested budget outside the range of your points is extrapolation: check it in steps.
- Channel revenue depends on attribution: under last-click attribution the return is overstated.