Average ROAS vs marginal ROAS is the gap between overall return and the next dollar’s return. This matters because diminishing returns make marginal ROAS fall as spend rises, so a 3x average doesn’t guarantee a 3x return. Without tracking that number, marketers underinvest judging a channel on weak dollars, or overspend past zero. This video explains why spend continues until marginal cost equals marginal revenue, not until average looks good.
Why Average ROAS vs Marginal ROAS Is the Real Budgeting Question
Average ROAS vs marginal ROAS is the real budgeting question, because a channel’s overall return and the return on its next dollar are almost never the same number once diminishing returns are in play. In this conversation, Dr. Dominique Hanssens explains why averages only work when a response is a straight line, comparing a $1,000 touchpoint against a $10,000 one on average return unfairly penalizes the larger spend, since it’s absorbing a steeper decline the smaller one never faced. He contrasts this with a stock investment, where returns are identical regardless of position size, precisely because marketing, unlike a stock price, is genuinely non-linear.
Hanssens argues the fix is to stop asking about the return on your dollar and start asking about the return on your last dollar, whatever the current spend level actually is, and to keep spending until marginal cost equals marginal revenue, the exact point where marginal return hits zero and profit is maximized. John Wallace adds the practitioner’s view: averages are easy to calculate, but they don’t tell you what slope of the curve you’re on, still climbing steeply or already tapering. He flags conversion lift tests specifically as a common trap, since they report a real average incremental ROAS, but marketers often mistake that average for what the next dollar will return, when by definition the next dollar is buying a less favorable position on the curve than the dollars already spent.
In this video, you’ll learn:
Why average marketing returns are often misleading
The impact of diminishing returns on budget allocation
How to identify the optimal point for ad spending
The limitations of standard conversion lift tests
Why you should shift focus toward marginal returns
TIME STAMPFull Video Transcript: Average ROAS vs Marginal ROAS Explained
+–
Dr. Dominique Hanssens and LiftLab CEO John Wallace explain why marketing decisions should be based on marginal ROAS, the return on the next dollar, not average ROAS.
[0:00 – 0:50] The Problem with Averages in Marketing
[Dr. Dominique Hanssens]: If the effects of marketing were a straight line, then averages would be just fine, because whether you measure the first, the second, or whatever, the average is always the same because the response is a straight line. Unfortunately, that’s not true, for reasons that we just discussed. There are these diminishing returns. What does that mean? It means, first of all, that you have to stop looking at these averages, because they’re not comparable. If you have one campaign, or pick one touchpoint, and you spend $1,000, and you pick another touchpoint and you spend $10,000, and then you compare the results, you can’t do that, because the second, you’re actually going to bias against the second one because you’re spending much more. You have to go away from the averages.
[0:50 – 1:56] Why Managers Misinterpret Investment Returns
[Dr. Dominique Hanssens]: It’s typically problematic in management for the following reason: for all the right financial reasons, managers like to talk about returns. For example, if we both invested in the same stock, we buy at the same day for the same price, let’s say $100. You buy 1,000 shares, I buy 100 shares. A year later we sell at the exact same price, let’s say $120. Your return for your thousand shares is exactly the same as mine with my 100 shares. The size of the investment doesn’t matter. In marketing, it does, because of the diminishing returns. You should expect a higher return from a small investment than from a large investment, and that, of course, can lead companies to underinvest at a certain standard, but they don’t realize that cutoff point for returns slides down as you invest more. That creates some very serious problems in misallocation.
[1:57 – 3:30] The Importance of the Marginal Return
[Dr. Dominique Hanssens]: What you really should be doing is not looking at the return on your dollar, but look at the return on your last dollar. I don’t mean here literally the last, it could be the last $100, the last thousand, but the highest point that you’re spending. Is that still a metric that is sufficiently high to justify the expense? You spend until your marginal cost is equal to your marginal revenue, which is the point where your marginal return is going to be exactly zero. In other words, you’re indifferent between spending more or spending less, because you’re exactly at the zero point. That is exactly at the point where you make the most money. In the methods that John talked about with LiftLab, the combination of historical data analysis and experiments, we draw these curves in very reliable ways. Now we find exactly the point where the spending no longer pays for itself. That is a big change in thinking: go away from these average return metrics to return of the last dollar, in order to make the appropriate decision.
[3:31 – 4:52] The Gap Between Averages and Diagnostic Curves
[John Wallace]: The challenge here is you need the curves. The averages are easy, because anybody can calculate one, but you sort of need to know what slope you’re on. Are you on a deep, very strongly increasing slope, where another million dollars may do very well, or is it starting to taper off? That’s where companies like LiftLab do such a good job pointing to exactly where you are with respect to each marketing investment, each touchpoint, each media. If I can take Mike’s comment and turn it into what I see run by marketers, we all struggle with some of the same bad revenue data, revenue data that’s reported on last click or as reported by an ad platform. Typically the first anecdote or correction of that would be, “let’s run a conversion lift test,” but that conversion lift test is going to give you an average. At that average spend during the experiment, this was my performance, but it isn’t giving you the curves that we’re referring to. Even though it’s aspirational, let’s run a lot of conversion lift tests, they’re still going to fall short of teaching you, “am I near the inflection point of this curve?”
[4:53 – 6:05] Avoiding Common Forecasting Mistakes
[John Wallace]: Just taking the exact statements from Dr. Hanssens and mapping them over to what we see as practitioners, those tests that do get you the average incremental performance are better than nothing. I endorse them as a step in the right direction, but they are going to fall short of giving you this level of diagnostic. Maybe the most common mistake that I see made, and it took me probably a couple meetings with Dr. Hanssens for this to sink in personally, is to see someone run an incrementality test, come back with a number like the average incremental ROAS is three, and make the conclusion that because of this test, if I spend another dollar, it will return three. That’s just not true. It’s very tempting to take that result and drop it into a spreadsheet and run a forecast, but you’re going to miss your forecast. Because when you spend the extra dollar, let’s say during the experiment you were spending $1,000 a day, when you bump your spend up to $1,100, you don’t get to buy that very early performance again. You’ve already committed to it with the first thousand, so you’re buying something that’s not working quite as well. By definition, you are going to see some return less than three. It’s the curves that teach you how below three this additional spend is going to perform.
Key Lessons: Shifting From Average ROAS to Marginal ROAS
Marketing analytics: Marketing impact is non-linear; relying on aggregate averages obscures performance differences across varying investment levels.
Budget strategy: Companies should aim to spend until marginal cost equals marginal revenue, maximizing profit at the point where marginal return hits zero.
Decision making: Managers must shift from viewing average ROAS to analyzing the performance of the most recent dollar spent, to avoid under-investment or misallocation.
Practical measurement: While conversion lift tests are useful, they only provide an average; practitioners need data curves to understand the slope of performance and accurately forecast future results.
