Strong campaigns can still waste budget. Here’s why.
Average metrics tell you how your campaign performed. Marginal metrics tell you what the next euro will do. Here's why that distinction matters more than most marketers realise.
Most marketers are working toward one of two things: growing faster or spending smarter. Either you have growth targets to hit and need every euro to pull its weight, or you’re under pressure to improve efficiency and make the same budget go further. The goals are different but the underlying question is the same: is the budget actually working as hard as it could be?
Average performance can mask a quieter problem: campaigns that seem strong overall are quietly becoming less efficient at the margin. The first euros spent were doing real work. The last euros? Much less so.
This isn’t an edge case. It’s how diminishing returns work: the more you spend, the less each extra euro delivers. It happens in almost every campaign that scales.
The gap between average and marginal
Average ROAS divides total revenue by total spend. It’s a useful number for reviewing what happened. But past performance doesn’t tell you where to put tomorrow’s budget.
That’s what marginal metrics measure. Marginal ROAS, POAS, CPA and CPC all answer the same question: if you spend one more euro on this campaign right now, what do you get back?
The answer is almost never the same as your average. The gap between the two grows the more you spend, to the point where a campaign with a strong average ROAS might be generating very little return on the last euros you added.
What diminishing returns actually look like
Take a campaign with a strong average ROAS. Zoom into the spend curve and each additional euro delivers less return than the last. The graph below shows how that decline progresses as spend increases.
The average ROAS across all of that spend might still look healthy. Nothing in the dashboard flags a problem. But you’ve been pushing budget into a campaign where the marginal return has dropped to near break-even.
The same pattern plays out with CPA and CPC. Early spend reaches the most convertible audience at the lowest cost. As you scale, you reach progressively harder-to-convert users, and the cost per acquisition climbs steeply.
Why average metrics keep you one step behind
The issue isn’t that average metrics are wrong. They’re accurate for what they measure. The problem is using them to decide where to spend next.
When you increase a campaign budget based on a strong average ROAS, you’re betting that the next euro will perform like the average of all the euros before it. It won’t. You’re spending into the diminishing end of the curve, where efficiency has already dropped.
By the time your average metrics reflect the inefficiency, you’ve already overspent. Marginal metrics give you the signal earlier, when it still matters. And the budget you free up doesn’t disappear. It moves to where the next euro still delivers, getting more out of the same budget.
What marginal metrics don’t replace
Average metrics still matter. A lot. They’re the right tool for performance reviews, for reporting to clients or stakeholders, and for understanding how a campaign has performed over time. That context doesn’t go away.
Marginal metrics are specifically a budget allocation tool. They answer a narrow but important question at the moment you’re deciding where to put the next euro. Outside of that decision, average metrics remain a clear and intuitive way to communicate performance.
The goal isn’t to replace one with the other. It’s to use each for what it’s actually good at. Average metrics tell you if a campaign was good. Marginal metrics tell you if you should spend more.
Marginal metrics in Billy Grace automations
Billy Grace now supports Marginal ROAS, POAS, CPA and CPC as optimisation targets inside automations. Instead of adjusting budgets based on historical averages, the model evaluates what the next euro is likely to return and acts accordingly.
Set a marginal threshold and the automation increases spending where the next euro is expected to outperform it. It pulls back where it won’t, reallocating the budget to where it continues to generate strong returns.
Our best practices
We recommend setting your threshold low to start, giving the system room to learn your campaign dynamics before you push for efficiency gains. From there, you can raise it gradually as your confidence in the data grows.
Marginal metrics work best when you let them run. Give the system time to adapt and learn, monitor which campaigns are scaled up or down and adjust based on what the data is telling you rather than on assumptions.
