Marginal ROAS is the revenue you earn from the next slice of ad spend, divided by that extra spend. Where ordinary return on ad spend averages every pound across the whole budget, marginal ROAS looks only at the last pounds added, which is where most of the waste in a growing account hides.
How marginal ROAS works
Suppose a shop spends £2,000 a month on Google Shopping and records £10,000 in sales. Average ROAS is 5, or 500%. The owner raises the budget to £3,000 and sales climb to £12,000. The headline ROAS is now 4, which still looks healthy. But the extra £1,000 bought only £2,000 of extra revenue, so the marginal ROAS on that increase is 2.
That gap exists because of diminishing returns. The first pounds in any campaign go on the cheapest, most relevant searches: people typing your brand or an exact product name. As the budget grows, the platform has to buy dearer clicks, broader searches and less likely buyers to spend it. Each extra pound works a little less hard than the one before, while the average hides the decline because it still includes the strong early spend.
You can estimate marginal ROAS in a few ways: compare two stable periods at different budgets, read the curve in the bid simulator for a campaign, or run a controlled test that changes spend in one set of regions and not another. None of these is exact. Seasonality, stock and promotions all move the numbers, so treat any single estimate as a range rather than a fact.
Why it matters
Budget decisions are made at the margin. The usual question is not whether the whole account should exist but whether to add another £500 next month or move £500 from Search to Performance Max. Average ROAS cannot answer that, and marginal ROAS can.
It matters most when it falls below your break-even point. If your gross margin after delivery and payment fees is 40%, a pound of ad spend needs about £2.50 of revenue to pay for itself. A campaign averaging 4 could be spending its last few hundred pounds at a marginal ROAS of 1.5, losing money on every one of those sales while the dashboard looks fine. For a UK retailer, remember to work in revenue excluding VAT: a ROAS figure that counts the 20% VAT you pass to HMRC flatters every calculation.
Common mistakes
- Raising the budget because average ROAS is above target, without checking what the increase itself returned.
- Comparing a quiet month at low spend with December at high spend and crediting the budget for what the season did.
- Setting a target ROAS that matches the average you want rather than the marginal return you can afford.
- Ignoring returns and cancellations, which cut real revenue and hit the marginal sales hardest when broad traffic brings in less committed buyers.
- Reading platform-reported revenue as if every sale was caused by the ad. Some of it would have happened anyway, which is the question incrementality testing answers.
How to act on it
Start by working out your break-even ROAS from real margins, ex VAT. Then, before any budget rise, write down what you expect the extra spend to return, make the change in one step, hold it for long enough to cover your normal buying delay, and compare like with like. If the increase returns less than break-even, step back and put the money somewhere with more headroom.
Bear in mind that a target ROAS is judged on the average across a campaign, so raising the budget or lowering the target lets the bidding buy auctions that return less than the figure you set. Read the bid simulator before each change, and if margins vary by product, feed the bidding profit rather than revenue, through profit on ad spend, so the extra spend is judged on what it actually earns. This kind of budget testing is a routine part of my monthly Google Ads management.