Marketing budget allocation by marginal return, not average ROAS

Average return is total revenue from a channel divided by its total spend. Marginal return is the extra revenue from the next slice of spend, divided by that slice. The glossary entry on marginal ROAS puts it simply: the return on the next euro rather than the average. A channel can have a good average return while its marginal return sits below break-even.

Two other terms sit nearby, and they are easy to mix up.

• Our convention for MER: MER (marketing efficiency ratio) is revenue divided by all marketing cost, not just media spend. It does not split credit between channels.
• Our convention for blended ROAS: blended ROAS is revenue divided by media spend only.
• Incremental ROAS (iROAS): revenue the ads caused, divided by ad spend. A trade publication on iROAS notes that ordinary ROAS counts purchases that would have happened anyway².

Averages answer "how did we do?". Marginal return answers "what should we do next?". Budget decisions are the second question.

Because the easy customers come first. That is our view of the usual pattern, and it has a few causes.

• The best-fit audience is reached first. Extra spend reaches people who are less likely to buy.
• Repetition. The same people see the same ads more often.
• Overlap with other channels. More spend can claim sales another channel would have won.
• Fixed pool of demand. Some people were always going to buy.

The open-source PyMC-Marketing project describes modelling adstock, the carry-over effect of advertising over time, together with saturation, which captures diminishing returns³. The name differs by tool. The idea is the same: each extra euro earns less than the one before.

That does not mean every channel flattens at the same point. Some saturate early. Some hold up for a long time. You find out by measuring, not by assuming.

You do not need a full model to get a rough answer. Marketing-mix modelling (MMM) is one route. These are cheaper ones, roughly in order of effort.

1. Spend steps. Raise or cut a channel's budget in a clear step. Hold everything else steady. Compare the change in total revenue with the change in spend. Watch for promotions and seasonality, which can swamp the signal.
2. Controlled tests. Hold spend back in a few regions or change it in a few. Compare against matched regions. This is a cleaner read than a step, because the other regions share your calendar.
3. A saturation curve. Plot spend against results over time and fit a simple curve. It is a rough picture, but it shows where the curve bends.

Example (illustrative): a brand spends €50,000 a month on a channel and gets €200,000 of revenue. Average ROAS is 4.0. It raises spend by €10,000, and revenue rises by €20,000. Marginal ROAS on that slice is 2.0. The new average is €220,000 ÷ €60,000, about 3.7, so it still looks healthy. The next euro says 2.0.

Rough estimates from controlled spend changes beat having none. They are not precise, and you should say so. A range with a stated method is more useful than one confident number.

Our view: use a step or a test first, and keep a saturation curve as the way to show it. Move to a model when you have many channels and enough history.

Compare each channel's marginal return with your break-even, not with another channel's average.

Break-even ROAS is the lowest ROAS at which ads do not lose money on the orders they bring. The glossary notes that it is 1 divided by margin, and that sources differ on which margin to use. Our convention: use contribution margin, meaning revenue minus variable costs, before marketing. That includes product cost, shipping, payment fees and returns, and it leaves out marketing.

Example (illustrative): contribution margin before marketing is 40%. Break-even ROAS is 1 divided by 0.4, which is 2.5. The slice in the previous example returned 2.0. That is below break-even, so the extra €10,000 loses money, even though the channel's average ROAS is still about 3.7.

With that in hand, the plan is simple.

1. List channels with their best estimate of marginal return, with a range.
2. Mark which are above break-even and which are below.
3. Take budget from channels below it. Add to channels above it.
4. Move in steps, not in one jump.
5. Re-measure after each step. The curve moves.

Compare also with an incrementality view where you have one. A marginal figure built on platform-reported revenue inherits the platform's habit of counting sales that would have happened anyway. Our Break-even ROAS and contribution margin calculator gives you the threshold before you move money.

Not a channel league table. Show the decision.

• The threshold. Break-even ROAS from contribution margin, with the definition.
• The estimate. Marginal return per channel, with a range and how it was measured.
• The move. How much budget goes where, and what you expect, in margin.
• The method. Step, test or model, and how confident you are.
• The next check. When you will re-measure, and what would change the plan.

Finance wants the next euro in margin terms. Marketing wants the next euro in revenue terms. The honest bridge is break-even.