Why does average ROAS fail for planning?
Average ROAS answers a backward-looking question: across everything we spent on this channel, how much revenue did we record per pound? That is genuinely useful for reporting and reconciliation. The trouble starts when teams use it to decide where the next pound should go, because an average is a single number stretched across spend that performed very differently. Your first pounds bought the cheapest, highest-intent demand. Your last pounds reached colder, more saturated audiences. Blending them together produces a figure that flatters the weakest spend and disguises the strongest.
The consequence is systematic misallocation. Planning on blended averages over-invests in channels that harvest existing demand and under-invests in channels that create it, because the harvesting channels post the highest headline ROAS while contributing the least incremental growth. This is the same blind spot that makes incremental return on ad spend (iROAS) so important: both metrics exist to stop you optimising for an attractive average that does not survive the next budget decision.
What does marginal ROAS measure?
Marginal ROAS measures the efficiency of the next increment of spend, not the average of all spend so far. Formally, it is the slope of the spend-to-revenue response curve at your current budget: if a small increase in spend produces a proportionally smaller increase in revenue, your marginal ROAS is low even when the average still looks strong. It is the only ROAS figure that directly answers the question every media plan actually asks - “if I move more money here, will it pay off?”
Because it is forward-looking, marginal ROAS is the metric that should drive allocation. When the marginal return of one channel falls below the marginal return of another, budget should move - regardless of which channel has the prettier average. This is why ElenIQ treats marginal efficiency, not blended averages, as the unit of decision when you build a media plan around a fixed budget.
A budget allocation example
Consider a channel running at £40,000 a month and returning £240,000 in revenue - a tidy 6.0 average ROAS. On paper it is your best performer, so the instinct is to pour more in. But suppose the last £5,000 you added only produced £10,000 of revenue. The average is still 6.0, yet the marginal ROAS on that final tranche is just 2.0, and the next tranche will be lower still. The headline metric is celebrating spend you should arguably be pulling back.
Now compare a second channel sitting at a 4.0 average ROAS - visibly “worse” - but where the next £5,000 would still return £17,500, a marginal ROAS of 3.5. Allocating by average sends money to the 6.0 channel and leaves growth on the table. Allocating by margin moves the next pound to the channel returning 3.5 at the margin, because that is where incremental revenue is actually created. The right move only becomes visible once you stop reading the average and start reading the slope.
How saturation and scaling change marginal ROAS
The reason marginal ROAS and average ROAS drift apart is saturation. Paid channels do not respond to budget in a straight line; they follow a curve that rises quickly at first and then flattens as the most responsive audiences are exhausted. Each additional pound reaches a slightly less valuable impression than the one before, so marginal return declines smoothly while the average - weighed down by all that efficient early spend - falls far more slowly. Scale a channel hard enough and its marginal ROAS can cross below break-even while the average still reads comfortably above it.
These dynamics are captured mathematically by the Hill function, the saturation curve most response models use to describe diminishing returns. Fit that curve to your data and the marginal return at any budget is simply its slope at that point. We explain how these curves reveal the ceiling on profitable spend in our guide to how saturation curves predict paid media budget ceilings.
How ElenIQ forecasts next-pound efficiency
ElenIQ is built around the principle that budget decisions should be made at the margin, before the money is committed. Rather than reporting a single blended ROAS, it fits a response curve to your historic channel data and forecasts the marginal return at every budget level, so you can see exactly where the next pound stops being worth it on each channel. Move a slider and the platform re-solves the allocation, surfacing the point at which marginal returns equalise across channels - the mathematically efficient split for your budget.
That is what “forecast before you spend” means in practice. You can pressure-test a plan with the marginal ROAS calculator, model a full reallocation in the ad spend forecasting tool, and do it all without the brittle, error-prone workbooks that usually carry this analysis - a workflow we cover in media planning without spreadsheets.
Frequently asked questions
What is the difference between marginal ROAS and average ROAS?
Average ROAS divides all of a channel’s attributed revenue by all of its spend, producing a single blended figure that describes performance after the fact. Marginal ROAS measures the return on the next pound of spend - the additional revenue created by increasing the budget slightly from where it sits today. Average ROAS is a reporting metric; marginal ROAS is a planning metric.
Why does average ROAS fail for budget planning?
Average ROAS blends your most efficient early spend with your least efficient incremental spend into one number, so it hides the point at which extra investment stops paying off. A channel can show a strong average ROAS while its marginal ROAS has already fallen below break-even, which means scaling it would lose money even though the headline figure still looks healthy.
Can a channel look strong on average ROAS but weak at the margin?
Yes. This is the central reason planners separate the two metrics. A channel can appear strong on average ROAS while becoming weak at the margin, because early spend captured cheap, high-intent demand and inflated the blended average, while the next pound now reaches saturated, lower-value audiences. The average looks reassuring; the margin tells you whether to add budget.
How is marginal ROAS related to iROAS?
Marginal ROAS and iROAS both describe causal, forward-looking efficiency rather than blended history. iROAS isolates the incremental revenue a channel truly creates versus what would have happened anyway, while marginal ROAS describes how that incremental return changes as spend rises along the response curve. Sound budget decisions use both: incrementality to set the true baseline, and marginal efficiency to find the ceiling.
How do you calculate marginal ROAS?
Marginal ROAS is the slope of the spend-to-revenue response curve at your current budget: the extra revenue produced by a small increase in spend, divided by that increase. Because the curve flattens as a channel saturates, marginal ROAS falls as you scale. ElenIQ fits a response curve to your historic data and reports the marginal return at each budget level so you can see exactly where the next pound stops being worth it.