Marginal ROAS

Marginal ROAS is the revenue the next euro of ad spend brings, as opposed to the average return of all euros spent so far. Because returns diminish as budgets grow, it sits below average ROAS and decides whether more spend pays off.

Daniel Busch
Written by Daniel Busch · Chief of Staff · Updated

In short

  • Formula: change in revenue ÷ change in ad spend between two budget levels
  • Falls as budgets grow (diminishing returns), so it is lower than the average ROAS every report shows
  • Scale a campaign while its marginal ROAS stays above your break-even ROAS
  • Estimate it with budget steps, holdout or geo tests, or the response curves of a marketing mix model
  • No benchmark exists, ours included: it belongs to one account, one budget level and one moment

How marginal ROAS works

Marginal ROAS = change in revenue ÷ change in ad spend

Average ROAS divides all revenue by all spend. Marginal ROAS looks only at the last step: what the extra euros brought. The two differ because returns diminish. A campaign’s first euros reach the people most likely to buy; each further euro reaches people less ready to buy, shows your ads to the same people more often and pays more for each impression. Meta’s open-source marketing mix model Robyn describes this as each additional unit of advertising increasing the response “at a declining rate”, and calls marginal ROAS “the most important metric for budget allocation” (Robyn).

An example on net revenue, with the break-even ROAS of 2.36 from the break-even ROAS example, a CM2 margin of 42.36%:

Daily budgetRevenueAverage ROASMarginal ROAS of the stepContribution after ads per day
€1,000€3,0003.00(starting point)€271
€1,500€4,0002.672.00€194
€2,000€4,6002.301.20−€51

At €1,500 a day, the average ROAS of 2.67 is still above break-even, yet the campaign earns €77 a day less than at €1,000, because the extra €500 brought only €2.00 per euro. At €2,000, the average of 2.30 has only just slipped below break-even, while the last €500 returned 1.20 and the day ends in a loss.

Why average ROAS hides it

Every ROAS in a report is an average: the revenue of a period over its spend. The average mixes the best first euros with the worst last ones, so it falls slowly while the marginal return falls fast, as in the example. Scaling decisions are about the last euros, which the average never shows.

Our benchmarks are averages too. Each brand’s ROAS in our Meta Ads ROAS benchmark is its attributed revenue over its spend for a month. No benchmark for marginal ROAS exists, ours included: it belongs to one account at one budget level at one moment, and it only shows when spend changes. A “typical marginal ROAS” quoted anywhere is an average under another name.

The same goes for moving budget between channels. A search channel’s average often includes brand search, which barely scales, because nobody searches for your brand more often when you bid more. So ask what the next €1,000 earns in each channel; only a budget step or a test tells you.

How to estimate marginal ROAS

MethodHowGood forLimits
Budget stepsRaise or cut one campaign’s budget by a clear step, hold everything else, compare revenue and spend with the period beforeShould we scale this campaign?Seasons and other changes blur it; needs a stable attribution model
Holdout or geo testStop or raise spend for a random group or a set of regions and compare with the restIs the channel incremental at all, and what does a step add?Takes weeks, costs revenue in the holdout, needs volume
Response curvesA marketing mix model fits revenue as a curve of each channel’s spend; its slope at your budget is the marginal returnSplitting the budget across channelsNeeds a long history with varied spend; a model, not a measurement

Google’s open-source mix model Meridian reports marginal ROI as “the incremental outcome on your next dollar spent above the historical budget level”, and reads a marginal ROI much lower than the average ROI as a sign that a channel “is beginning to saturate” (Google).

Budget steps are the method every brand can run this month. Our advice for reading them:

  • Make the step large enough to see, 20 to 30% of the budget, and keep it for one to two weeks. Daily swings are noise.
  • Use click-date allocation, so each order lands on the day of the click that led to it and the step’s revenue is not mixed with orders from clicks before the change.
  • Change one thing at a time. A new creative or a promotion during the step makes it unreadable.
  • Let the learning phase settle after a big budget change before you read the step.
  • Use the same attribution model and revenue basis as your break-even ROAS.

What is a good marginal ROAS?

Any marginal ROAS above your break-even ROAS, on the same revenue basis and attribution model: the extra spend then pays for itself out of contribution margin. The best budget is where the marginal ROAS of the next step falls to break-even, not where the average does. In the example that point lies between €1,000 and €1,500 a day; steering by the average would have taken the budget to almost €2,000.

Our view: compare channels on their marginal returns, never on their averages, and read prospecting channels on first touch or on new-customer revenue, as you would their average ROAS. A channel with a lower average can still have the better next euro.

FAQ about Marginal ROAS

What is the difference between marginal ROAS and average ROAS?

Average ROAS divides all revenue by all spend; marginal ROAS divides the extra revenue of a budget step by the extra spend. Because returns diminish, marginal ROAS is lower than average ROAS while you scale, and it is the one that tells you whether more spend pays off.

How do you calculate marginal ROAS?

Take two budget levels and divide the change in revenue by the change in spend. If raising a campaign from €1,000 to €1,500 a day lifts revenue from €3,000 to €4,000, the marginal ROAS of the step is €1,000 ÷ €500 = 2.0, while the average ROAS is still 2.67.

Is marginal ROAS the same as incremental ROAS?

No. Incremental ROAS compares revenue with and without the ads; marginal ROAS compares two budget levels. A geo or holdout test of a budget step measures both at once: the incremental revenue of the extra spend.

Why does ROAS fall when I raise the budget?

Diminishing returns: the extra budget reaches people less likely to buy and shows your ads to the same people more often. A learning phase after a big budget change on Meta or Google adds short-term noise, so read a budget step once it has settled.