You measure marketing ROI beyond ROAS by measuring the business instead of the ad account. That means four numbers rather than one: a blended return across all marketing cost, a return calculated on gross profit rather than revenue, a payback period on what you paid to acquire each customer, and an incrementality read that tells you what would have happened anyway. ROAS is one input into those numbers. On its own it is a platform's report card on itself.
This matters more than it sounds. A 3.0 ROAS campaign can lose money once you count cost of goods and agency fees, and a campaign reporting half that can be the most valuable spend in the business. The number on the dashboard does not tell you which situation you are in, and most founders find out eighteen months late.
Nielsen's 2025 research puts a number on how common the gap is. 85% of marketers say they are confident in their ability to measure ROI, and only 32% measure it across both their traditional and their digital channels. Confidence is not the scarce resource here, and measurement is.
What is the difference between ROAS and marketing ROI?
ROAS is revenue divided by ad spend, calculated inside one platform, on revenue that platform believes it caused. Marketing ROI is profit divided by total marketing cost, calculated on your own numbers, across everything you spent.
Those are different questions with different answers, and the gap between them is where businesses quietly lose money.
| ROAS | Marketing ROI | |
|---|---|---|
| Numerator | Revenue the platform attributes to itself | Gross profit, after cost of goods |
| Denominator | Media spend in that account | Media, agency fees, tools and salaries |
| Who supplies the data | The platform selling you the ads | Your finance system |
| Question it answers | Did this ad account produce revenue? | Did marketing make the business money? |
| Where it fails | Ignores margin, ignores every other cost | Slower to read, needs clean cost data |
ROAS does its own job well, and that job is comparing one campaign against another inside the same account under the same conditions. It is a steering wheel. The mistake is treating it as the speedometer, the fuel gauge and the profit and loss statement as well.
Why does ROAS overstate what your ads earned?
Three reasons, and all three have got worse since 2021.
The first is that the platform marks its own homework. Meta counts a conversion when someone clicks your ad within seven days or merely views it within one day, then buys. Somebody who was already searching for your brand, saw your retargeting ad in passing and bought anyway gets counted as a Meta conversion. The revenue on that line is real, and the causation behind it is an assumption the platform is making about its own work.
The second is that the tracking itself broke. When Apple introduced App Tracking Transparency in 2021, Meta told investors on its Q4 2021 earnings call that the change would cost it roughly $10 billion of revenue in 2022. The lost revenue was the visible half of that. The other half was the signal Meta stopped receiving about who bought what, and platforms across the industry responded by modelling the conversions they could no longer observe. A growing share of what your dashboard reports is now an estimate rather than a record.
The third is that the industry's default attribution logic got simpler, not smarter. Google now supports only two attribution models in Ads and Analytics, having removed first-click, linear, time-decay and position-based in 2023. Whatever your reporting says, it is now built on last click or on Google's own data-driven model. Google built both of those to allocate credit inside Google, and neither one sets out to tell you what your marketing was worth to your business.
Why does ROAS also understate what your marketing earned?
Here is the part most performance-only agencies skip. ROAS is wrong in both directions at the same time.
Most of your future customers are not buying this month, so nothing you do to them shows up in this month's ROAS. Professor John Dawes of the Ehrenberg-Bass Institute set out the 95:5 rule for the LinkedIn B2B Institute in 2021: at any given moment, around 95% of potential buyers in a category are not in the market at all. Dawes is explicit that the ratio is a heuristic rather than a precise figure, and the point holds regardless. The overwhelming majority of the people your advertising reaches cannot buy today, which means the work of being remembered is invisible to any window shorter than the buying cycle.
Then there is everything your ads caused that they never got credited for. Somebody sees your ad, searches your brand name three weeks later and converts through organic. Somebody screenshots it and sends it to a friend who buys. Your search traffic rises during a heavy paid month and falls when you go quiet, and none of that reaches the ad account. If you have ever wondered why the brand strategy you paid for never showed up in the ads, part of the answer is that the scoreboard could not see it, so nobody defended it.
The four numbers that tell you whether marketing is working
Each number answers a question the one above it cannot. Run all four and you can tell the difference between an account that is working and a business that is growing.
| Number | How to calculate it | What it tells you |
|---|---|---|
| Blended MER | Total revenue ÷ total marketing cost | Whether marketing as a whole pays for itself |
| Contribution-margin ROI | (Gross profit − marketing cost) ÷ marketing cost | Whether it makes money after cost of goods |
| CAC payback period | Acquisition cost ÷ gross profit per customer | How long your cash is tied up per customer |
| Incremental ROAS | Incremental revenue ÷ ad spend | What would not have happened without the ads |
Blended MER is the one to start with, because it needs no new tooling. Take every ringgit of revenue the business booked, divide it by every ringgit you spent on marketing including agency fees, tools and the salary of whoever runs it, and track that ratio month over month. It cannot tell you which channel worked. It can tell you whether the whole operation is getting more efficient, which is the question the ad account is structurally unable to answer.
A worked example in ringgit
Here is an illustrative month for a Malaysian direct-to-consumer brand with a 45% gross margin. The numbers are constructed, and the shape of them is the one we see most often.
| Line | Amount |
|---|---|
| Meta ad spend | RM40,000 |
| Agency retainer, tools and in-house salary share | RM18,000 |
| Total marketing cost | RM58,000 |
| Revenue Meta attributes to itself | RM120,000 |
| Actual revenue booked by the business | RM148,000 |
| Gross profit at 45% | RM66,600 |
| Marketing contribution after all costs | RM8,600 |
The ad account reports a 3.0 ROAS, which reads like a healthy month. Blended MER is RM148,000 over RM58,000, or 2.55. Contribution-margin ROI is RM8,600 over RM58,000, which is a 14.8% return on the marketing budget. Read as a return, a 3.0 ROAS implies 200%. The business kept 14.8%.
At a 45% gross margin, break-even ROAS is 1 divided by 0.45, which is 2.22. So this account is running 35% above break-even on media alone and roughly at the line once the retainer is included. You can check your own break-even figure with our break-even ROAS calculator.
Now add incrementality. Suppose a holdout test shows that 25% of the revenue Meta claims would have arrived anyway through brand search and repeat purchase. Incremental revenue drops to RM90,000, and incremental ROAS on the RM40,000 of media is 2.25 against a break-even of 2.22. The campaign is not losing money. It is also nowhere near the three-to-one machine the dashboard describes, and the founder scaling on that 3.0 figure is scaling on the wrong number.
One more read from the same month. If those RM148,000 came partly from 480 new customers, acquisition cost is RM58,000 over 480, or roughly RM121 per customer. At an average order value of RM185 and a 45% margin, the first purchase returns about RM83 of gross profit. The first order does not repay the acquisition. The business is being funded by the second one, which makes retention a marketing metric rather than a customer service one.
Can you run an incrementality test without a data team?
Four options, from crudest to cleanest.
The pause test is free and blunt. Turn one channel off for two weeks, hold everything else steady and watch total revenue rather than the channel's own reporting. If total sales barely move, that channel was harvesting demand you already had. It is a rough instrument and seasonality can mislead you, and it will still tell you more than another month of dashboard staring.
Geo holdouts are the practical middle ground. Split Malaysia into matched regions, run ads in one set and withhold them in the other, then compare total sales between the two. Klang Valley against Penang and Johor is usually close enough to be readable for a national brand.
Platform lift tests do the same thing with better statistics. Meta's conversion lift studies hold back a randomised share of your audience and measure the difference in conversion rate, which removes the seasonality problem. They need meaningful volume and a few weeks to run, so they suit accounts spending consistently rather than in bursts.
Marketing mix modelling used to be the preserve of companies with analysts on staff. Google released Meridian, its open-source Bayesian MMM, to everyone in January 2025, and its documentation is explicit that incrementality experiments are the strongest basis for calibrating the model. That ordering is the useful lesson even if you never run a model. Experiments first, modelling second, dashboards last.
Frequently asked questions
What is a good marketing ROI?
There is no universal figure, because it depends on your gross margin. Start from break-even instead. Divide 1 by your gross margin to get the ROAS at which media pays for itself, then include agency fees, tools and salaries to find the point at which all marketing pays for itself. Above that line you are making money, and below it you are subsidising your own growth out of working capital.
Should I stop tracking ROAS?
No. ROAS is the right tool for comparing campaigns and creatives inside one account under the same conditions, and it reads fast enough to make daily decisions on. Keep it as the steering input. Move the budget decisions to blended MER and contribution-margin ROI, which are the numbers that reflect the business.
How often should I measure incrementality?
Once or twice a year for most Malaysian SMEs, and after any significant change in spend, channel mix or positioning. Incrementality tests cost you real revenue while the holdout is running, so they are not a monthly exercise. The insight tends to hold for a couple of quarters.
What is MER, and how does it differ from ROAS?
MER, or marketing efficiency ratio, is total business revenue divided by total marketing spend. It counts every ringgit of revenue regardless of which channel gets credit, and every ringgit of cost including agency fees and tools. ROAS counts only what one platform believes it caused, against only that platform's media cost. MER is harder to game and harder to misread.
How do I measure the ROI of brand marketing?
Not with a conversion window. Track branded search volume, direct traffic, share of voice and the conversion rate of your paid campaigns over six to twelve months, then watch whether your cost per acquisition falls as those rise. Brand work shows up as a discount on everything else you buy, which is why it is invisible to any channel-level report.
Can I do this with separate agencies for each channel?
It is difficult, because blended numbers need someone accountable for the total rather than for one account. When four vendors each report their own ROAS, every conversion gets claimed more than once and nobody owns the blended number underneath. We covered how that fragmentation compounds in our guide to building a marketing system that compounds.
The bottom line
ROAS tells you what happened in an ad account. Marketing ROI tells you what happened to your business, and those two stories diverge most sharply in the months when the dashboard looks best. Start with blended MER, because you can calculate it this afternoon from numbers you already have. Add margin. Run one holdout test this year.
If you are not sure which of your numbers is the real one, talk to us and we will work through your actual figures with you, whether or not you end up working with us. You can also read our Malaysian Meta ads cost and ROAS benchmarks to see where your account sits against the market, or look at how our performance marketing team reports on the business rather than the platform.