What is ROAS?
ROAS, or return on ad spend, is the revenue a channel generates divided by what was spent on it:
ROAS = Revenue from Ads ÷ Ad Spend
That looks like a fixed calculation, but the revenue side depends on which measurement method calculated it, so the same channel can carry a different ROAS depending on whether last-click, an MMM, or a total commerce view produced the number. A ROAS figure is really the output of several moving parts, so when one of those parts shifts, the reported number moves even when nothing about the ads themselves changed.
What causes a ROAS shift?
Four things can explain a ROAS shift, and the topline number alone won't tell you which one it is.
Attribution and measurement bias
Fospha's 2024 analysis, based on comparing last-click-attributed conversions against Fospha's incrementality-based measurement across its client base, found Meta drove 226% more conversions than last-click attribution credited it with, and TikTok drove 788% more, rising to 702x for TikTok's Awareness and Consideration activity alone.
Last-click attribution assigns a sale to whichever click sat closest to the purchase, so it systematically undervalues upper-funnel activity, like display or paid social, that built the demand a search or retargeting ad later captured. When more of a channel's real contribution goes uncounted, its reported ROAS drops even though the channel got no less effective. For example, a brand running YouTube alongside Brand Search can see this play out. YouTube builds the demand, Search gets the click, and the channel doing the heavy lifting ends up looking weak on a last-click ROAS.
Cross-channel and marketplace measurement gaps
A 2024 Amazon Ads case study found that 64% of purchases attributed to a leading apparel brand's campaigns happened off Amazon entirely, with a third of those off-platform sales part of a broader halo effect: a sale one channel drives on another sales surface, like a TikTok ad prompting a purchase on Amazon. It's a single-brand result Amazon published to showcase its own measurement tools, but the pattern shows up across other retailers too.
A channel builds the demand, a sale happens, but the measurement system was never looking in the right place to count it. When that happens, the channel that created the sale shows a lower ROAS than it actually earned. Learn more about the halo effect in retail media.
Diminishing returns from scaling spend
Spend more on a channel, and eventually each extra dollar buys less than the last one did, because there's a finite pool of people ready to buy at any given time. For example, a brand that doubles its Meta Advantage+ budget hoping to double revenue might instead see spend up 100% but incremental revenue up closer to 40%, because the campaign is already reaching most of the audience likely to convert.
This is called diminishing returns, and marketing mix models usually plot it as a saturation curve, a line that flattens out as spend climbs. ROAS falls even though the channel is still working, just less efficiently at the new spend level.
Channel interdependency
Google's own Brand Lift data found that eHealth saw a 45% increase in branded search volume after running a display campaign, demand that shows up in Search's numbers, not Display's. The example is dated, but the mechanic, budget in one channel moving another channel's numbers, still shows up in marketing mix models today.
Budget decisions in one channel can also move another channel's reported ROAS, with nothing about that channel actually changing. Cut a Display budget, and Brand Search ROAS can shift even though Search's own performance stayed flat, because Display was the one quietly building the branded demand Search had been capturing.
How do you know what caused the ROAS shift?
Diagnosing which of these is at play means looking at the shift from a few different angles rather than reading the blended number alone. Teams can sort a ROAS movement into four categories:
CarParts.com, a Fospha client, ran into this in practice with a Meta ROAS movement that looked like a performance decline. Running a Conversion Lift Study to check which category was driving it, they confirmed Meta was still generating incremental revenue, then scaled Meta investment by over 200% year-over-year, making it their top channel for ROAS.
Fospha's platform can automate this diagnosis, analyzing performance across a full marketing portfolio and flagging which of the four categories is driving a shift, so a team starts with a structured explanation instead of reconstructing one from exports.
How do you fix a ROAS shift once you know the cause?
Whichever category explains the shift, treat it as a starting point for investigation. A portfolio mix shift during a promotional week, or a revenue variance around a seasonal spike, can look identical in the numbers to a genuine performance problem. Validate the diagnosis against seasonality, promotions, and any strategic budget decisions made during the same period before moving any budget.
This is where Fospha comes in. It runs attribution and MMM daily, giving teams always-on signal between incrementality tests, and flags when a test like the one above is worth running to confirm what the data is suggesting.
Common questions
Q: Does a ROAS drop always mean a problem?
Not necessarily. A drop can reflect a measurement gap, like an uncounted halo effect or a channel hitting its saturation point, rather than a real decline in what marketing is contributing. Checking which of the four diagnostic categories is driving the number, and validating it against business context, is what tells the difference between a real problem and a reporting artifact.
Q: Can more than one of these causes happen at the same time?
Yes. A spend increase that pushes a channel past saturation can happen in the same week as a seasonal revenue swing, and the two effects can move ROAS in opposite directions at once. That's why isolating each contributing factor separately, rather than reading the blended number alone, matters for an accurate diagnosis.
Q: Is a rising ROAS always a good sign?
Not always. A channel can show a higher ROAS simply because spend was cut and it's now serving only the highest-intent audience left, not because it became more effective. A rising number is worth checking against spend and revenue trends before treating it as evidence of improved performance.
Related reading
- What is Total Commerce ROAS and how is it different from platform ROAS?
- What are the 5 questions to ask an MMM provider before signing?
See how Fospha works in practice. 30-minute walkthrough on your data, your channels, this quarter.
