MER vs ROAS: The Third Number Both Metrics Hide From You

Marketing Efficiency Ratio (MER) is total revenue divided by total marketing spend. Return on Ad Spend (ROAS) is the revenue one platform attributes to itself divided by what you spent on that platform. The difference matters. Almost every article about it stops one step too early. Both numbers are averages of what already happened, and neither tells you where your next dollar should go. A third number answers the question both of them dodge.

MER vs ROAS: The Difference in One Paragraph

ROAS is a channel diagnostic. It answers a narrow question: for the money spent inside this ad account, how much revenue did this platform claim? MER is a business scoreboard. It answers a wider question: for every dollar the company spent on marketing, how much total revenue came in? A brand spending $100,000 across all marketing and generating $500,000 in revenue has an MER of 5. That same brand might see Meta report a 4.0 ROAS on its own spend while Google reports 6.0. The three numbers can all be accurate and still describe different things, which is why teams that treat them as interchangeable end up arguing about which dashboard is lying.

One note on the name. MER is sometimes written as Media Efficiency Ratio and sometimes as Marketing Efficiency Ratio. Marketing Efficiency Ratio is the more accurate expansion, because the denominator includes agency fees, content production, influencer contracts and tooling, not only media. It is also occasionally called blended ROAS, which is a reasonable shorthand and a slightly misleading one, since the word ROAS carries the attribution baggage MER was meant to escape.

Why ROAS Misleads: Four Failures That Compound

ROAS has four structural problems, and they compound.

The Platform Grades Its Own Homework

Every ad platform decides for itself which conversions it caused. Meta counts a view-through. Google counts a Performance Max impression. AppLovin counts a checkout that followed an in-app video. None of them can see the other touchpoints, and none of them have any incentive to claim less. This is the same structural problem that breaks multi-touch attribution as a budget tool: a model that only sees tracked touchpoints cannot tell you what would have happened without them. Add the reported returns across every platform and you routinely get more revenue than the business booked.

ROAS Rewards Harvesting Over Creating

A retargeting campaign shown to people already on their way to buy will post a spectacular ROAS. It did not create that revenue. A prospecting campaign that introduces the brand to someone who buys six weeks later will post a poor one. Optimizing to ROAS pushes budget toward the campaigns that document demand and away from the campaigns that generate it.

ROAS Ignores Customer Lifetime Value

A campaign with a 3.0 ROAS that acquires subscribers who stay two years is worth more than a 7.0 ROAS campaign selling one-time discounted orders. ROAS closes its books at the first transaction. Customer lifetime value (CLV, sometimes LTV) is the number that captures the difference, and no ad platform can see it, because the platform's view of the customer ends at checkout.

ROAS Counts Revenue, Not Contribution Margin

ROAS divides revenue by spend. Revenue is not money you keep. A 3.0 ROAS on a product carrying 30% gross margin returns 90 cents of gross profit per dollar of ad spend. That is a loss before a single fixed cost is counted. Contribution margin, revenue minus cost of goods and variable costs, is the version of the number that tells you whether the campaign made the company money. Two campaigns at identical ROAS can sit on opposite sides of break-even if their product mix differs, and neither the platform nor MER will show you that. This is why finance and marketing so often read the same dashboard and reach opposite conclusions: one is looking at revenue efficiency and the other at margin.

The metric survives its flaws for one reason. It is fast, it updates hourly, and it sits in the interface where the buying decision gets made.

Why MER Does Not Fix the Problem ROAS Created

Most comparisons of these two metrics stop here. This is where the argument gets interesting.

MER solves the attribution problem by refusing to play. It ignores which platform claimed what and compares total revenue to total spend, so no amount of double-counting inside the ad platforms can inflate it. That is a real improvement, and it is why finance teams like it.

MER also cannot tell you anything about a channel. MER is one number for the entire business. If it moves from 4.2 to 3.8 across a quarter, you know something changed. You do not know whether Meta saturated, a competitor entered the auction, your email list aged, your best creative burned out, or you simply spent more in a slower month. The metric that cannot be gamed is also the metric that cannot be acted on.

Worse, MER moves for reasons that have nothing to do with marketing. Launch a product, run a promotion, get a press hit, hire a good merchandiser, and MER improves while marketing did nothing differently. A rising MER can mean the marketing got better or it can mean the business got better around the marketing. From the number alone you cannot tell which.

A paid media lead at a D2C apparel brand described the version of this that shows up in practice. The marketing mix model reported improving efficiency. The team's own internal numbers said the opposite. Two credible measurement systems, two directions. Adding a third top-line ratio does not break the tie.

Both Numbers Are Averages, and Budgets Get Decided at the Margin

ROAS and MER are both averages. That is the shared flaw underneath the surface disagreement, and it is the reason tracking both does not resolve anything.

An average return describes the money you already spent. It answers "how did this pool of spend perform overall?" No budget decision is ever that question. The actual question is always incremental: if I move $50,000 from Meta to YouTube next month, what happens to revenue? A blended 4.0 average says nothing about the return on the fifty-thousand-and-first dollar. The first dollar into a channel and the millionth dollar into that same channel do not perform remotely alike.

Channels saturate. Early spend reaches the people most likely to convert. Later spend reaches people who are progressively less likely to, and at some point an additional dollar produces less revenue than it costs. A channel can hold a healthy average ROAS long after its marginal return has gone underwater, because the strong early dollars are still propping up the mean.

This is the practical failure. A team reviewing a dashboard of averages cannot see which channel is the best home for the next dollar. So the split defaults to last quarter's, plus or minus a nudge.

Marginal ROAS: The Number That Answers the Budget Question

Marginal ROAS, sometimes written as mROAS or discussed as marginal ROI, is the return on the next dollar rather than the average return on all dollars spent so far. It is the number that actually maps to a budget decision.

The distinction has teeth. A channel running a 5.0 average ROAS with a 0.8 marginal ROAS should be cut, because every further dollar is losing money even though the channel looks like a winner. A channel running a 2.5 average with a 3.1 marginal should be scaled, because it is still climbing. Read the averages alone and you would do the opposite of the correct thing in both cases.

Getting this number honestly takes a model. The common substitute is a year-over-year trend comparison. An analytics lead at a subscription marketplace was blunt about his team's method before working with us. Marginal return came off year-over-year trends, which he acknowledged was not really data science. That is what is available when nothing better exists.

Marginal return cannot be read off a platform dashboard, because the platform does not know what would have happened without it. It has to be estimated from a model of how each channel responds to spend, and that model has to be anchored to something causal.

How Causal Measurement Produces a Number You Can Spend Against

Two methods produce marginal return honestly, and they work best together.

Bayesian Marketing Mix Models Estimate the Response Curve

Bayesian marketing mix models estimate how revenue responds to spend across every channel at once, including offline and unattributable ones, using aggregate data rather than user-level tracking. Channel-specific adstock and saturation curves capture the fact that each channel decays and saturates on its own schedule. Because the models are Bayesian, they produce uncertainty intervals instead of false-precision point estimates, so you can see how confident the read on any given channel actually is.

Geo-Based Incrementality Tests Observe the Counterfactual

Geo-based incrementality tests turn a channel off in a set of matched regions, then compare the regions where it ran against a synthetic counterfactual built from the holdout markets. This is the closest thing to a controlled experiment available in live media. It answers the causal question directly: what would sales have been without this spend?

The two reinforce each other. Test results update the model's priors, and the model tells you which channel is worth testing next. That loop is what turns a measurement exercise into a budget process. At BlueAlpha this runs in the decision layer, and the model refits weekly rather than quarterly. That cadence matters, because a marginal return estimate arriving ninety days late describes a media landscape that no longer exists.

What Cann and beehiiv Found When They Tested Their Reported Numbers

Two published tests show the size of the gap between reported and real.

Cann, a cannabis beverage brand selling direct to consumers, could not get a straight answer on AppLovin. The platform's own dashboard swung between 2x and 16x ROAS on any given day. Triple Whale, the third-party tool the team treated as the neutral referee, showed 4.77 for December.

A 19-day geo holdout across five states, measured against Shopify transactions rather than modeled conversions, found no statistically significant lift at Cann's $10K-a-week spend level. The point estimate was positive at $90K. The 95% credible interval ran from negative $152K to positive $310K, which is another way of saying the effect could not be distinguished from zero. Cann paused the channel and moved roughly $40K a month into Google, reallocating about $480K a year.

Across the test window AppLovin's own dashboard reported 0.43x, while the independent third-party tool reported 4.77. The tool brands trust as the neutral referee was the one furthest from what the holdout measured. Blending is not the same thing as accuracy.

For beehiiv, a newsletter platform, geo tests across four channels found Meta's true incremental signup CPA was roughly 345% higher than Meta reported, at 93% confidence. The same tests found TikTok's incremental CPA was about 10% lower than reported, at 99% confidence. The four channels tested were Meta, TikTok, YouTube and LinkedIn; LinkedIn returned an inconclusive read at 60% confidence because conversion volume during the window was too low. Platforms err in both directions, and a blended number cannot tell you which way any individual channel is wrong. beehiiv right-sized Meta and scaled TikTok and YouTube on the evidence.

Neither of those findings is reachable from an average. You can stare at MER and ROAS side by side for a year and never learn that AppLovin was contributing nothing.

Running the Budget Like a Portfolio Instead of a Scoreboard

One version of the marketing leader spends the quarter assembling evidence that last quarter's budget was defensible. Screenshots from four platforms, a blended number that finance accepts, a story that holds until someone asks a harder question. The metrics debate is a symptom of that posture: when you cannot prove what a dollar did, you argue about which proxy to trust.

A portfolio manager does not run that way. They know the return on each position, they know which position deserves the next dollar, and they rebalance continuously because the proof is in hand. They are not defending last quarter. They are deploying this one.

That shift is what causal measurement buys, and it is why the metrics question is smaller than it looks. MER and ROAS are both fine as diagnostics. ROAS tells you whether an ad account is functioning. MER tells you whether the business is growing faster than the marketing budget. Keep both on the dashboard. Just stop expecting either to tell you where the money should go, because that is not a question an average can answer.

Where to Start When Your Metrics Disagree With Each Other

If your platform numbers and your internal numbers point in different directions, the useful next step is not choosing a side. It is running one causal test on the channel where the disagreement is most expensive, and using the result to anchor everything else. That is usually the largest channel nobody can defend, or the one whose reported returns look too good to be true.

BlueAlpha builds the model and runs the tests, then wires the result into the ad accounts so the decision reaches the platform instead of a deck. We start with the channel you argue about most. Book a channel performance audit and we will show you what your reported numbers are hiding.

Proof points in this article come from BlueAlpha's published case studies. Cann's geo holdout ran 19 days from January 13 to 31, 2026, across five holdout states. It measured Shopify transaction data using a Bayesian structural time series model with a synthetic control. beehiiv's tests ran approximately six weeks across four channels. Reported ROAS and CPA figures are quoted as the source systems reported them. Practitioner comments are drawn from client calls and are anonymized.

FAQ

What is the difference between MER and ROAS?

ROAS is the revenue a single ad platform attributes to itself divided by the spend on that platform. MER, or Marketing Efficiency Ratio, is total company revenue divided by total marketing spend across every channel. ROAS is attribution-dependent and channel-specific. MER ignores attribution entirely and describes the whole business. It is also called blended ROAS, though MER usually includes agency fees, content and tooling, while blended ROAS is sometimes limited to media spend.

How do you calculate Marketing Efficiency Ratio?

Divide total revenue for a period by total marketing spend for the same period. A company generating $500,000 in revenue on $100,000 of marketing spend has an MER of 5.0. Many teams also calculate a new-customer MER, using only revenue from first-time buyers, to separate acquisition efficiency from retention.

What does it mean when an agency says they optimize for MER instead of ROAS?

It usually means they are optimizing to a number their own reporting cannot inflate, which is a genuine improvement over channel-level ROAS. It also means they have chosen a metric that cannot isolate their contribution from everything else moving your revenue: promotions, product launches, seasonality, press. Ask which channels they can prove are incremental and how they proved it. If the answer is the MER trend line, they have not answered the question.

Should I use ROAS or MER for Facebook and Meta campaigns?

Use ROAS inside the account for day-to-day optimization decisions such as which creative and audience to scale. Use MER at the business level to check whether total marketing is producing growth. Neither tells you whether Meta specifically is incremental. That requires a geo holdout test or a marketing mix model that isolates Meta's causal contribution. BlueAlpha's tests for beehiiv found Meta's true incremental signup CPA was roughly 345% higher than the platform reported.

What are the limitations of using ROAS to measure marketing success?

ROAS relies on the platform's own attribution, which systematically over-claims credit. It rewards retargeting that harvests existing demand over prospecting that creates it. It ignores customer lifetime value by closing at the first transaction. It counts revenue rather than contribution margin, so a strong ROAS can still lose money at low gross margin. And as an average, it says nothing about the return on the next dollar of spend.

What is marginal ROAS and why does it matter more than average ROAS?

Marginal ROAS is the return on the next dollar spent in a channel rather than the average return across all dollars already spent. It matters because channels saturate: a channel can hold a strong average while its marginal return has already fallen below break-even. Budget decisions are marginal decisions, so marginal return is the number that maps to them.

My MMM and my attribution model disagree. How do I reconcile them?

Do not average them, and do not pick the one you prefer. Run a geo holdout test on the channel where they disagree most and treat the measured lift as ground truth. Then recalibrate whichever model was further off. An incrementality test is the only one of the three that observes a counterfactual rather than modeling one.