Marginal ROAS
The revenue you get back from the next rupee of ad spend, not the average rupee. It's the number that tells you whether adding budget will make money or just make revenue.
Measured between two spend levels held over comparable periods, or read off a response curve from a marketing mix model.
What is marginal ROAS?
ROAS is an average. It divides all the revenue a campaign drove by all the money it spent. But the first ₹10,000 of a day's budget goes to the people most likely to buy, and each extra rupee reaches people a little less likely to. Returns fall as spend rises.
Marginal ROAS looks only at the last slice of spend. Because the average still includes all those cheap early sales, it can look healthy long after the last rupee has started losing money. When you're deciding whether to scale, the last rupee is the one that matters.
How it's calculated
No ad platform reports marginal ROAS. You get it by comparing two spend levels, or by fitting a response curve to your spend and revenue history.
| Input | Where it comes from | What it gives us |
|---|---|---|
| spend | Ad platform, average per day at each budget level | The change in spend |
| attributed_revenue | True (not platform) attribution, same window at each level | The change in revenue for that campaign |
| response_curve | Marketing mix model fitted on daily spend and sales | Slope of revenue vs spend at any budget |
How to measure it
There are two practical routes. Spend-step tests are simple and campaign-specific. A marketing mix model covers every channel at once but needs a longer history.
- 1Hold a baselineRun the campaign at its current budget for 5 to 7 days and record true ROAS.
- 2Step up onceRaise budget by 20 to 50% and hold it for another 5 to 7 days. Change nothing else.
- 3Divide the deltasExtra daily revenue ÷ extra daily spend is the marginal ROAS for that step.
- 4Repeat or modelKeep stepping while it clears break-even, or use MMM response curves to see every level at once.
Worked examples
Example 1: Scaling one campaign in steps
An illustrative prospecting campaign is stepped from ₹50,000 to ₹1,25,000 a day, one week per level. The brand's break-even ROAS is 2.0.
| Daily spend | Daily revenue true attribution | Average ROAS | Marginal ROAS |
|---|---|---|---|
| ₹50,000 | ₹1,75,000 | 3.50 | baseline |
| ₹75,000 | ₹2,40,000 | 3.20 | 2.60 |
| ₹1,00,000 | ₹2,85,000 | 2.85 | 1.80 |
| ₹1,25,000 | ₹3,10,000 | 2.48 | 1.00 |
- ₹75,000The extra ₹25,000 brought back ₹65,000. Marginal ROAS of 2.6 clears break-even, so this step was worth it.
- ₹1,00,000Average ROAS still reads 2.85, comfortably above 2.0. But the extra ₹25,000 returned only ₹45,000, a marginal ROAS of 1.8. That slice is losing money.
- ₹1,25,000The last ₹25,000 brought back ₹25,000. The average of 2.48 still looks fine on a dashboard. The right budget for this campaign is around ₹75,000 a day.
Example 2: Two campaigns, same average, different headroom
Both illustrative campaigns show an average true ROAS of 3.0. You have ₹20,000 a day of extra budget to place.
Using marginal ROAS to decide
Compare marginal ROAS to your break-even ROAS, never to the campaign's own average. Use these bands as a starting point and tune them to how aggressively you want to grow.
Common mistakes
- Scaling on average ROAS"ROAS is still 2.8, push more" is how campaigns end up overspent. Check the last step.
- Testing over different weeksComparing a sale week to a normal week measures the sale, not the spend.
- Using platform-reported revenuePlatforms claim more credit as spend grows. Use true attribution on both sides of the step.
- Jumping too far at onceDoubling budget overnight resets learning and hides where returns started to drop.
Frequently asked questions
Is marginal ROAS the same as incremental ROAS?
No. Incremental ROAS asks whether sales would have happened without the ad at all. Marginal ROAS asks what the next rupee adds on top of what you already spend. You can have both: the incremental return of your next slice of budget.
Can I get marginal ROAS without running tests?
A marketing mix model can estimate it from your spend and sales history, as long as spend has varied enough over time for the model to see the curve.
Why does it fall as I spend more?
Ad platforms deliver to the people most likely to convert first. As budget grows, they reach further into less interested audiences and show ads more often to the same people, so each extra rupee does a bit less.
How long should each spend step run?
Long enough to cover your attribution window and a full week of weekday patterns. Five to seven days per step is a reasonable place to start.
Related metrics
Scale on the next rupee, not the average
BooleanMaths gives you true attribution to measure spend steps on, and marketing mix modelling to see response curves across channels.
Marginal ROAS
The revenue you get back from the next rupee of ad spend, not the average rupee. It's the number that tells you whether adding budget will make money or just make revenue.
Measured between two spend levels held over comparable periods, or read off a response curve from a marketing mix model.
What is marginal ROAS?
ROAS is an average. It divides all the revenue a campaign drove by all the money it spent. But the first ₹10,000 of a day's budget goes to the people most likely to buy, and each extra rupee reaches people a little less likely to. Returns fall as spend rises.
Marginal ROAS looks only at the last slice of spend. Because the average still includes all those cheap early sales, it can look healthy long after the last rupee has started losing money. When you're deciding whether to scale, the last rupee is the one that matters.
How it's calculated
No ad platform reports marginal ROAS. You get it by comparing two spend levels, or by fitting a response curve to your spend and revenue history.
| Input | Where it comes from | What it gives us |
|---|---|---|
| spend | Ad platform, average per day at each budget level | The change in spend |
| attributed_revenue | True (not platform) attribution, same window at each level | The change in revenue for that campaign |
| response_curve | Marketing mix model fitted on daily spend and sales | Slope of revenue vs spend at any budget |
How to measure it
There are two practical routes. Spend-step tests are simple and campaign-specific. A marketing mix model covers every channel at once but needs a longer history.
- 1Hold a baselineRun the campaign at its current budget for 5 to 7 days and record true ROAS.
- 2Step up onceRaise budget by 20 to 50% and hold it for another 5 to 7 days. Change nothing else.
- 3Divide the deltasExtra daily revenue ÷ extra daily spend is the marginal ROAS for that step.
- 4Repeat or modelKeep stepping while it clears break-even, or use MMM response curves to see every level at once.
Worked examples
Example 1: Scaling one campaign in steps
An illustrative prospecting campaign is stepped from ₹50,000 to ₹1,25,000 a day, one week per level. The brand's break-even ROAS is 2.0.
| Daily spend | Daily revenue true attribution | Average ROAS | Marginal ROAS |
|---|---|---|---|
| ₹50,000 | ₹1,75,000 | 3.50 | baseline |
| ₹75,000 | ₹2,40,000 | 3.20 | 2.60 |
| ₹1,00,000 | ₹2,85,000 | 2.85 | 1.80 |
| ₹1,25,000 | ₹3,10,000 | 2.48 | 1.00 |
- ₹75,000The extra ₹25,000 brought back ₹65,000. Marginal ROAS of 2.6 clears break-even, so this step was worth it.
- ₹1,00,000Average ROAS still reads 2.85, comfortably above 2.0. But the extra ₹25,000 returned only ₹45,000, a marginal ROAS of 1.8. That slice is losing money.
- ₹1,25,000The last ₹25,000 brought back ₹25,000. The average of 2.48 still looks fine on a dashboard. The right budget for this campaign is around ₹75,000 a day.
Example 2: Two campaigns, same average, different headroom
Both illustrative campaigns show an average true ROAS of 3.0. You have ₹20,000 a day of extra budget to place.
Using marginal ROAS to decide
Compare marginal ROAS to your break-even ROAS, never to the campaign's own average. Use these bands as a starting point and tune them to how aggressively you want to grow.
Common mistakes
- Scaling on average ROAS"ROAS is still 2.8, push more" is how campaigns end up overspent. Check the last step.
- Testing over different weeksComparing a sale week to a normal week measures the sale, not the spend.
- Using platform-reported revenuePlatforms claim more credit as spend grows. Use true attribution on both sides of the step.
- Jumping too far at onceDoubling budget overnight resets learning and hides where returns started to drop.
Frequently asked questions
Is marginal ROAS the same as incremental ROAS?
No. Incremental ROAS asks whether sales would have happened without the ad at all. Marginal ROAS asks what the next rupee adds on top of what you already spend. You can have both: the incremental return of your next slice of budget.
Can I get marginal ROAS without running tests?
A marketing mix model can estimate it from your spend and sales history, as long as spend has varied enough over time for the model to see the curve.
Why does it fall as I spend more?
Ad platforms deliver to the people most likely to convert first. As budget grows, they reach further into less interested audiences and show ads more often to the same people, so each extra rupee does a bit less.
How long should each spend step run?
Long enough to cover your attribution window and a full week of weekday patterns. Five to seven days per step is a reasonable place to start.
Related metrics
Scale on the next rupee, not the average
BooleanMaths gives you true attribution to measure spend steps on, and marketing mix modelling to see response curves across channels.