ROAS vs MER vs Incrementality: The Real Metric

Open Ads Manager and you see one number bigger than the rest: ROAS. It feels like the verdict on your whole account. It is not. Platform ROAS is the most misread metric in performance marketing, because it counts sales Meta merely watched, not sales Meta caused. This article takes apart the three metrics that actually decide whether your advertising makes money: ROAS, MER and incrementality. We use real 2025 numbers and real cases. The goal is simple. By the end you will stop steering by a number that flatters you, and start steering by the number that pays you. The difference between those two habits is, for many accounts, the difference between scaling and slowly bleeding.

What ROAS actually measures, and what it hides

ROAS, return on ad spend, is revenue attributed to your ads divided by what you spent. Simple maths, treacherous word: attributed. Meta credits itself with any sale where the buyer saw or clicked an ad inside the attribution window. That includes people who were already going to buy. A customer who searched your brand, found you on Google, then happened to scroll past a retargeting ad gets logged as a Meta conversion. Meta did not create that sale. It witnessed it. The ROAS in your dashboard bundles caused sales and coincidental sales into one flattering figure, and you cannot tell them apart from inside the platform.

There is a structural reason this number inflates. Self attributing networks, the platforms that grade their own homework, average an 18% duplicate attribution rate across the industry. Meta and Google both claim the same conversion, so the sum of your platform ROAS overcounts reality. The closer a campaign sits to the bottom of the funnel, the worse it gets. Retargeting is the extreme case, because it targets people already in your orbit. We will see numbers shortly where reported retargeting ROAS of 8x collapses to 2x once you strip out the sales that would have happened anyway. The platform is not lying on purpose. It is structurally built to over credit itself.

The myth: a high ROAS means a profitable campaign

Kill this one first. A high ROAS does not mean you made money. ROAS is built on revenue, and revenue is not profit. A 4x ROAS on a product with a 25% contribution margin can lose you money once you load in cost of goods, shipping, payment fees, returns and discounts. Meanwhile a 2x ROAS on a 70% margin digital product is comfortably profitable. The number that decides survival is not how much revenue an ad returns, it is how much margin survives after every variable cost. Reading ROAS without your margin attached is like reading a speed without knowing the speed limit. The same figure can be a triumph or a disaster depending on a context the metric itself never shows you.

Contribution margin: the floor under every ROAS target

Before you set any ROAS target, you need your break even ROAS, and that comes from contribution margin. Contribution margin is revenue minus all variable costs: cost of goods, shipping, fulfilment, payment processing, returns. Note the word all. The most common mistake is using gross margin, which only subtracts cost of goods and ignores the rest. Gross margin tells you a product looks healthy. Contribution margin tells you whether you can afford to advertise it. Your break even ROAS is simply one divided by your contribution margin percentage. At 40% contribution margin, you break even at 2.5x. Below that, every extra euro of spend burns cash, no matter how proud the dashboard looks.

This is why POAS, profit on ad spend, is gaining ground in 2025. POAS optimises on actual gross profit instead of gross revenue, folding product margins, shipping and fees into the calculation. A practical way to make Meta bid on profit: push contribution margin values into the platform through the Conversions API or value rules, so the system optimises toward margin, not turnover. The single most common error here is the same one again, feeding gross margin instead of contribution margin. Get the input right and a target ROAS quietly becomes a target POAS, which is what you actually wanted all along. The platform will happily chase profit if you give it profit to chase.

Margins set the stakes, so anchor your thinking in them. DTC operators in 2026 are working with contribution margins that vary wildly by category, and a textbook profitable acquisition can run at a first order POAS below 1x when lifetime value carries it. One worked example from the field: a customer whose lifetime contribution margin reaches 85 pounds is profitable even at a 12 pound acquisition cost, despite a first order POAS that looks like a loss. Read that twice. A campaign that appears to lose money on the first sale can be your best investment once repeat purchases land. ROAS sees none of this, because it only ever looks at the first transaction inside its window.

MER: the number your CFO trusts

MER, marketing efficiency ratio, is brutally simple: total revenue divided by total marketing spend, across every channel, over a period. No attribution, no pixels, no windows. It cannot lie about which channel did what, because it never claims to know. That is its strength. While platform ROAS asks did this pixel say it worked, MER asks for every euro we spend on marketing, how much revenue do we get back. It maps directly onto your bank account, which is exactly why finance teams trust it more than any in platform number. MER is the metric that survives contact with your profit and loss statement, and that is a higher bar than any dashboard clears.

The gap between MER and platform ROAS is where the truth lives. One DTC brand saw a 2.0 ROAS inside Meta Ads Manager but a 3.5 MER across the business, revealing the halo effect of paid social on organic and email sales. Here the platform understated Meta, the opposite of the usual story, because Meta drove demand that closed elsewhere. The lesson is not that ROAS always lies high or always lies low. It is that ROAS and MER measure different things, and only by watching both do you see whether your paid spend is creating revenue or just relabelling it. One number alone will always mislead you in one direction or the other.

What a good MER actually looks like in 2025

Benchmarks help, as long as you read them against your margins. In 2025, the median MER across Triple Whale customers landed around 2.4, roughly a 41% marketing cost as a share of revenue. A healthy target sits between 3.0 and 5.0, meaning three to five euros of revenue per euro of total marketing spend. High margin businesses can survive at 3.0, while thin margin ecommerce often needs 4.0 or more just to stay profitable. Notice how that mirrors the contribution margin logic exactly. A good MER is not a universal figure. It is whatever number clears your break even once every cost is in, which is the same test you apply to ROAS.

MER has one real blind spot, and honest operators name it. Because it ignores attribution entirely, MER cannot tell you which channel to scale or cut. If your MER drops, you know the machine is less efficient, but not why. Was it Meta, Google, a creative slump, a price change, seasonality. MER is a thermometer, not a diagnosis. That is precisely why you do not pick MER instead of ROAS. You run them together: MER as the truth check on whether the whole engine is profitable, channel ROAS as the rough steering wheel, and incrementality as the calibration that connects the two. Each one fails alone and works as part of a set.

Incrementality: the only metric that proves causation

Incrementality answers the question ROAS dodges: how many of these sales would not have happened without the ad. The method is experimental, not statistical guesswork. You split your audience or your geographies into a test group that sees the ad and a control group that does not, then measure the difference in conversions. That difference is the incremental lift, the sales your advertising actually caused. Everything else in the test group, the people who would have bought anyway, is not credit you earned. It is credit Meta took. Incrementality is the only approach that separates the two with evidence rather than assumption, which is why it is the calibration layer for everything else.

The numbers from 2025 are sobering, and they are the heart of this article. Stella ran 46 incrementality studies between January and April 2025 across ecommerce brands doing 15 to 100 million dollars a year, using geo based inverse holdout testing. One documented case from Times One Hundred logged 481 total attributed purchases but only 171 incremental, a 64% gap between what Meta claimed and what it caused. A fitness apparel company spending 100,000 dollars a month at a reported 4:1 ROAS found a true incremental ROAS of 2.5:1, meaning 37% of attributed revenue would have arrived without the ads. These are not edge cases. They are the rule once you measure instead of trusting the dashboard.

Retargeting: where the overstatement is worst

Retargeting is the cleanest illustration of the problem, because it targets people already primed to buy. One documented WooCommerce store saw Meta report an 8x ROAS on retargeting, while incremental attribution revealed that 75% of those conversions would have happened anyway. That 8x reported ROAS was a 2x incremental ROAS. The customers were already heading to checkout. The ad just got in front of them and claimed the win. This is not an argument to kill retargeting outright, it has its place, but to stop scaling it on a number that is three quarters fiction. Most accounts are wildly over invested in retargeting precisely because its ROAS looks unbeatable on the surface, when in reality it is the one place the inflation runs deepest and the wasted budget is hardest to spot.

Meta finally admits it: incremental attribution

Meta itself moved on this in 2025, which is the clearest sign the old ROAS was overstated. In April 2025 Meta introduced incremental attribution inside Ads Manager, and by 19 May 2025 it was available to all advertisers through Ads Manager, the Marketing API and third party platforms. Instead of crediting every click, it estimates which conversions were truly caused, drawing on Meta’s accumulated lift study data and machine learning. You enable it in the columns settings, compare attribution settings, then check incremental attribution under advanced options. For the first time, the platform shows you a deflated, more honest number next to the flattering one, side by side in the same report, so you no longer have to guess which one to trust.

Meta told investors that advertisers using the feature were seeing an average 46% lift signal in incremental conversions, framing it as a feature, but read what it implies. If incremental attribution materially changes the picture, then the standard attribution it sits beside was materially overstating. There is also a real benefit beyond honesty: when you optimise toward incremental conversions, the algorithm stops chasing people who would have bought anyway and starts hunting genuine net new demand. The deflated number is not bad news. It is the number you should have been bidding against all along, and now the system can bid against it for you automatically. Turning it on costs nothing and changes nothing about your spend, it simply replaces a flattering column with a defensible one that you can take to a finance meeting without flinching.

New customers versus existing customers

Here is the distinction that reconciles everything. Most overstated ROAS comes from crediting Meta with sales to people who already know you. Real growth comes from net new customers, and that is exactly where Meta tends to be genuinely incremental. A 2025 Measured analysis of over 10,000 campaigns found Meta drives strong net new customer acquisition, and that 64% of the incremental conversions Meta produces come from new to brand customers. So the platform is most honest precisely where it matters most: bringing you buyers you did not already have. The trick is to measure and optimise for that segment, not the blended blob that quietly mixes acquisition with cheap repeat purchases.

This is why new customer acquisition cost, nCAC, beats blended CAC for steering. nCAC is prospecting spend divided by first time customers only. The gap is huge: one worked example showed a blended CAC of 75 dollars that became 180 dollars once filtered to new customers alone. If you steer by the blended figure, you flatter yourself with cheap repeat buyers and starve genuine acquisition. CAC has climbed roughly 60% over five years, with median DTC brands spending well over 100 dollars per new customer in 2026, while Meta CPMs rose about 20% in 2025. In that environment, knowing your true cost to acquire a stranger is not optional, it is the difference between growth and stagnation.

How to actually pilot: a metric stack, not a single KPI

Stop looking for the one true metric. There isn’t one. The operators who win in 2025 run a stack where each metric checks the others. MER sits at the top as the profitability truth, the number that has to clear your contribution margin break even for the whole business to make money. Underneath, platform ROAS gives you fast, granular steering inside Meta, fully aware that it overstates. Incrementality calibrates ROAS into reality. And nCAC plus lifetime value tell you whether the new customers you are buying will pay you back over time. Each metric covers another’s blind spot, and no single one of them is allowed to be the verdict on its own.

The practical loop looks like this. Run a periodic incrementality test, geo based if you can, to get your incrementality factor, the ratio of incremental conversions to platform reported ones. If Meta reports 500 conversions and your test shows 300 incremental, your factor is 0.6. Multiply your platform ROAS by that factor to get true ROAS. A 4x reported at a 0.7 factor is a 2.8x real return, and if your break even is 3x you are quietly losing money while the dashboard cheers. Refresh the factor each quarter, because it drifts as your mix, creative and audiences change. Between tests, steer day to day on platform numbers, but judge profitability on MER and on your true, calibrated ROAS rather than the raw one.

A worked example, end to end

Picture a homeware brand. Meta reports a 4x ROAS and the founder is delighted. Contribution margin is 35%, so break even ROAS is 2.86x. So far so good. Then a geo test returns an incrementality factor of 0.65, dragging true ROAS to 2.6x, below break even. The blended MER, meanwhile, sits at 3.1x, propped up by repeat email buyers Meta never touched. Three numbers, three different stories. The honest read: the paid engine is slightly underwater on incremental terms, kept afloat by retention. The fix is not to cut Meta blindly but to shift spend from over credited retargeting toward genuine prospecting, where Meta’s incremental lift is strongest.

Notice what each metric did. ROAS alone said scale up. MER alone said the business is fine, do nothing. Only the combination, calibrated by incrementality and read against margin, produced the right move: reallocate, do not cut, and chase net new demand. That is the entire argument of this article in one example. No single number is the verdict. The verdict emerges when you stack them and let each one expose where the others lie. The marketers who internalise this stop having the same circular argument about ROAS targets every quarter, because they have replaced one flattering number with a small system that tells the truth.

What this means for how you run Meta

Translate the theory into habits. First, set your break even ROAS from contribution margin, not gross margin, and write it down so every target references it. Second, feed margin values into Meta through the Conversions API so the algorithm bids toward profit, not turnover. Third, turn on incremental attribution in Ads Manager and watch the deflated number, not the inflated one. Fourth, separate new from existing customers and optimise prospecting for net new acquisition, where Meta is genuinely incremental. Fifth, review MER weekly against your break even as the final profitability check. None of this is exotic. It is just refusing to be flattered by a single screenshot.

The mindset shift is the real deliverable. Platform ROAS is a tool, not a truth. It is fast and granular and useful for steering inside the auction, but it systematically overstates your impact, worst of all on retargeting and existing customers. MER keeps you honest about the whole business. Contribution margin sets the floor everything has to clear. Incrementality is the calibration that turns a flattering number into a real one. Run them as a system and you will make fewer confident, expensive mistakes. Run on ROAS alone and you will keep scaling campaigns that look like winners and quietly drain your bank account, one over credited conversion at a time.

Sources

Triple Whale (MER definition, 2025 median MER and benchmarks); Northbeam and Eightx (MER versus ROAS, DTC benchmarks); Stella (46 incrementality studies, January to April 2025; Times One Hundred 481 versus 171 case); Haus (incrementality factor methodology, 640 experiments); Seresa and Myntagency (fitness apparel 4:1 versus 2.5:1, 18% duplicate attribution, WooCommerce 8x retargeting case); Measured (2025 analysis of over 10,000 campaigns, 64% of incremental conversions from new to brand customers); Meta investor communications and Ads Manager documentation (incremental attribution rollout April to May 2025, average 46% lift); JudeLuxe and Saras Analytics (POAS, contribution margin versus gross margin); Luca and admetrics (DTC margins and POAS examples); 27five and Prospeo (2025 to 2026 Meta ROAS, CPM and CAC benchmarks). Figures from agencies are reported by their authors and not independently audited by Meta.