Why ROAS 8 can be worse than ROAS 4

Two ROAS figures do not tell you which is better

ROAS 8 looks better than ROAS 4, but the business lives on the amount in euros that is left after advertising. A ratio can fall while that amount rises. Which figure is better cannot be read from the two numbers alone: it depends on the step in between, on your margin and on whether the additional revenue is real.

Suppose a shop with a 40% contribution margin rate before advertising faces this choice. All numbers here are invented and round; they show the mechanism, not an expectation.

ROAS halves and the shop earns €2,000 more. What exactly belongs in the rate, meaning goods, shipping, fees and returns, is covered in the article on costs in the contribution margin. If you calculate with the bare product margin, none of the following lines hold.

Marginal ROAS only measures the additional money

Google describes marginal ROAS as the additional value or revenue you get from additional spend. It is calculated as additional revenue divided by additional cost, in the example €80,000 on €30,000, so 2.67.

For a decision, the margin is still missing. A marginal ROAS of 1 means one additional euro of advertising brings one euro of revenue. At a 40% contribution margin, 40 cents of it come back, so you lose 60 cents.

A budget step pays off if: marginal ROAS > 1 ÷ contribution margin rate

The right-hand side is your break-even ROAS, here 2.5. But you compare it with the last step, not with the average. That is the difference from the usual reading of an account: the average also contains the first, cheap euros and therefore stays above break-even long after the last euro has started losing money. Google, too, writes that looking at average ROAS for individual segments can be misleading. And the bid target is itself an average: individual conversions may land above or below it as long as the overall picture fits.

The same numbers tip over at a 37.5% margin

The same pair of numbers is right for one shop and wrong for another. The tipping point is €30,000 ÷ €80,000 = 37.5%.

That is why I consider any ROAS comparison between shops or product ranges worthless without the margin. At 50%, the step brings €10,000 more; at 30%, it costs €6,000. Switching to POAS does not answer the question either: the margin is then included, but it is still a ratio, and it says nothing about additional cost and additional profit.

An intermediate step can beat both

Comparing only two points hides where the money is lost. Suppose the same shop reaches ROAS 6 at €20,000 of spend, so €120,000 in revenue. At a 40% contribution margin, the staircase looks like this:

ROAS 4 beats ROAS 8, and yet the last step burns €4,000, because its marginal ROAS of 2.0 is below the threshold of 2.5. The overall calculation of 2.67 had hidden that. Even at a 30% margin the middle step wins here with €16,000 against €14,000, although ROAS 4 clearly loses at that margin. So the question is rarely "8 or 4" but rather: at which budget step does marginal ROAS fall below your threshold?

Additional revenue only counts if it would not have come otherwise

The calculation above assumes that every reported euro is genuinely additional. If, at a 40% margin, a fifth of the additional revenue would have come anyway, €64,000 of real revenue remains, which means €25,600 of contribution margin against €30,000 of additional cost. The €2,000 gain becomes a €4,400 loss. In the example, a share of just over 6% that would have come anyway is already enough, because the lead was so narrow.

This also affects the high ROAS itself. Field experiments at eBay found returns from paid search that were a fraction of conventional estimates. Brand ads had no measurable short-term benefit there. For non-brand ads, new and infrequent users responded positively, while frequent users accounted for most of the spend without being influenced. eBay is extremely well known, so the order of magnitude cannot be transferred to a smaller shop. I still consider the direction plausible: high values often arise where the intent to buy was already there.

That gives a second way in which ROAS 8 can be worse, this time with the same budget. Suppose two campaigns each cost €10,000 at a 40% margin:

You only know these shares from a test, not from the account. How that works is covered under measuring incrementality.

First the threshold, then the experiment

First you set the threshold: 1 ÷ contribution margin rate, plus a deduction for the share that would probably have come anyway. At a 40% margin and an assumed 20% of sales that would have happened anyway, it is 1 ÷ (0.4 × 0.8) = 3.125 instead of 2.5.

For a first rough figure, the simulators in Google Ads are useful; in Search campaigns they estimate cost and conversion value under other settings. Take two neighbouring points and calculate the marginal ROAS between them. That remains an estimate based on reported values and does not replace a measurement.

Then you move the target in steps. According to Google, a target ROAS that is too high can limit traffic, and lowering the target gradually brings you into more auctions and more volume. According to Google, it reacts immediately but needs one to two conversion cycles to reach the new target. How large a step should be is covered in the article changing target ROAS.

The cleanest route is an experiment. Google recommends changing only one variable and defining one or two success metrics beforehand. I would use contribution margin after advertising in euros as the metric, calculated from your own numbers. ROAS falls in this test by design, so you cannot read success from it.

Scarce stock and scarce cash argue for the higher ROAS

The rule is my assessment and deliberately simplified. It leaves four things out of the calculation, three of which argue for the higher ROAS.

Capacity. The rule applies when you have stock and can deliver. If you are sold out or cannot restock, the higher ROAS is right.

Liquidity. In the first example, the step ties up €30,000 of ad spend for €2,000 of additional profit. That is just under 7% on the additional money, and the assumptions have hardly any buffer. If cash is tight or you can put it to better use elsewhere, skip the step.

Step-fixed costs. More orders can require staff, warehouse space or customer service. If that is not in your rate, the threshold is too low.

Repeat purchases. They work in the opposite direction. If the step brings in many new customers who buy again, even a marginal ROAS below the threshold can be justified. That is then a deliberate investment with its own calculation, see advertising below break-even on purpose.

How to explain this to the managing director

Leave out the ratio and talk about the step, in euros and in one sentence. For the middle step, for example: "For €10,000 more advertising, €40,000 of additional revenue came in, which is €16,000 of contribution margin, so €6,000 more is left. ROAS fell from 8 to 6 in the process. The next step to ROAS 4 would have cost €4,000, so I am not taking it."

So that the falling ratio does not dominate the debate again in next month's report, change the target metric: contribution margin after advertising as the headline figure, ROAS alongside it as a diagnostic. If you are the managing director yourself, do not ask your team or your agency about ROAS, but about the last budget step: what did it cost, how much contribution margin did it bring, and what shows that the revenue was additional?