ROAS vs ROI: Which Number Actually Tells You If Your Marketing Is Making Money
A Singapore retailer we spoke to had a very good year on paper. Their Meta account reported a 4.1x return on ad spend, month after month. The founder had the number on a whiteboard. Then the accountant closed the books and the business had made almost nothing on the products those ads were selling.
Nothing was broken and nobody had lied. The 4.1x was real, in the sense that the platform was measuring what it says it measures. It was just answering a question the founder was not asking. ROAS tells you how efficiently your media bought revenue. It says nothing at all about whether that revenue left any profit behind.
This guide sets out both numbers properly: the formulas, a full worked Singapore example including the two GST traps that quietly distort local calculations, how to derive the break-even ROAS your own business needs, the lead-generation version for service companies with no revenue in the platform, and where blended ROAS, MER and incremental ROAS fit.
It sits under our performance marketing guide for Singapore, alongside how to read your monthly marketing report.
The two formulas, and why they disagree
ROAS = attributed revenue ÷ ad spend. A 4x ROAS means every dollar of media was associated with four dollars of revenue. It is a media efficiency metric. Its denominator contains one cost, and its numerator contains no costs at all.
ROI = (gain from the investment − cost of the investment) ÷ cost of the investment. Expressed as a percentage. It is a profitability metric. Done properly, its numerator is gross profit and its denominator includes every cost of running the marketing, not just the media.
The gap between them is not a rounding error. It is the entire cost base of your business. ROAS is the speedometer; ROI is whether you can afford the journey.
A worked Singapore example
Take a Singapore e-commerce brand running one month of Meta and Google. The platforms report S$40,000 of revenue on S$10,000 of media — a clean 4.0x. Here is what happens when the same month goes through the P&L.
| Line | Amount (S$) | Note |
|---|---|---|
| Revenue reported by the platforms | 40,000 | Gross order value, as passed to the pixel |
| Less GST collected (9% inside the gross figure) | −3,303 | 40,000 ÷ 1.09 = 36,697 net. The GST was never yours. |
| Net revenue | 36,697 | The figure your margin applies to |
| Less cost of goods (55% of net) | −20,183 | A 45% gross margin |
| Less payment processing (~2.9%) | −1,064 | Card and wallet fees |
| Less delivery and packing | −2,400 | Averaged across orders |
| Less returns and refunds (4%) | −1,468 | Modest for the category |
| Contribution before marketing | 11,582 | What marketing has to be paid out of |
| Less media spend | −10,000 | The ROAS denominator |
| Less GST on Google media (if not GST-registered) | −270 | See the GST section below |
| Less agency fee, creative and tools | −3,200 | Excluded from ROAS entirely |
| Result | −1,888 | A 4x ROAS month that lost money |
That is the whole argument in one table. The campaign was efficient at buying revenue and unprofitable at generating it, and no amount of staring at the ROAS column would have revealed that.
Your break-even ROAS, in ten minutes
There is no good ROAS in the abstract. There is only the ROAS you need, and it comes straight out of your margin: break-even ROAS = 1 ÷ gross margin.
| Gross margin | Break-even ROAS (media only) | Realistic target once fees, returns and delivery are included |
|---|---|---|
| 20% | 5.0x | 6.5x or higher — paid acquisition is very hard at this margin |
| 30% | 3.3x | ~4.3x |
| 40% | 2.5x | ~3.2x |
| 50% | 2.0x | ~2.6x |
| 60% | 1.7x | ~2.1x |
| 70% | 1.4x | ~1.8x |
The right-hand column assumes roughly a quarter of contribution goes to the costs ROAS ignores; adjust it to your own numbers rather than treating it as a benchmark. The useful discipline is not precision, it is having a written line so that “4x sounds good” stops being a strategy.
A refinement worth knowing: strictly, you should build the break-even off contribution margin rather than gross margin — that is, gross margin after the variable costs that hit every order, such as payment processing, shipping and returns. It is always lower than gross margin, and it is the honest basis for the line.
The two Singapore GST traps
Trap one: GST inside your revenue
If your platform sends the gross order value to the pixel — which most default e-commerce setups do — then 9% GST is sitting inside the numerator of your ROAS. On a S$40,000 month that is roughly S$3,303 of revenue that was never yours. Your true ROAS is about 8.3% lower than the dashboard shows. Either configure the platform to pass net revenue, or apply the divisor before you compare against your break-even line.
Trap two: GST on your media, which depends on who you are
This one produces genuinely different answers for two businesses running identical campaigns:
- Google Ads has charged GST on Singapore-billed accounts since the rate rose to 9% on 1 January 2024. If you are GST-registered, that is recoverable as input tax and is a cash-flow matter rather than a cost. If you are not registered, it is a real 9% surcharge on all your search media that never appears in the ROAS calculation.
- Overseas suppliers under the Overseas Vendor Registration regime operate differently: a GST-registered Singapore business that supplies its GST registration number to a GST-registered overseas supplier should not be charged GST on those digital services in the first place.
Two consequences. First, check whether your GST number is actually on file with each platform — it is a two-minute job that is very often left undone. Second, if you are below the registration threshold, remember that your true media cost on Google is 9% above the reported figure, which pushes your required ROAS up accordingly. This is not tax advice; confirm your own position with your accountant or against IRAS guidance.
If you sell services, ROAS is the wrong tool entirely
Most Singapore SMEs are service businesses. There is no revenue event in the ad platform, because the sale happens weeks later over a phone call. Chasing a ROAS number here means either fabricating conversion values or ignoring the question. Do neither — work backwards instead.
A worked example for an aircon servicing and installation company:
- Average installation job: S$4,200 revenue, 38% gross margin → S$1,596 gross profit per job.
- Of raw leads, 55% are qualified (real person, in Singapore, wants the service you sell).
- Of qualified leads, 22% become jobs.
- So one raw lead is worth 0.55 × 0.22 × S$1,596 = S$193 in gross profit.
That S$193 is the absolute ceiling on cost per raw lead — the point at which the marketing generates exactly zero for the business. If you want marketing to contribute half its gross profit back to the company, your target cost per lead is about S$96, and your target cost per qualified lead is about S$175. Now the monthly report has a line to be judged against, and “our cost per lead is S$140” becomes a statement with a meaning.
Published Singapore cost-per-lead benchmarks do exist, generally quoted in the S$45–S$130 range across professional sectors, but they are agency-compiled estimates rather than audited data and vary enormously by industry. Use your own arithmetic as the target and treat any published range as a sanity check only. Our guide to lowering cost per lead on Meta covers the levers once you know your number.
Four flavours of “return”, and when each is the right one
| Metric | Formula | The question it answers | When to use it |
|---|---|---|---|
| Platform ROAS | Attributed revenue ÷ that platform’s spend | “Which campaigns and audiences inside this account are working?” | Daily and weekly optimisation. Never as a board number. |
| Blended ROAS | Total revenue ÷ total paid media spend | “Is the whole paid engine efficient?” | Weekly sanity check. Immune to double-counting between platforms. |
| MER (marketing efficiency ratio) | Total revenue ÷ all marketing cost | “Is the whole marketing function paying for itself?” | Monthly, at management level. Includes fees, tools and salaries. |
| Incremental ROAS | Additional revenue caused ÷ spend | “What would not have happened without the ads?” | Quarterly, via a holdout test. The only version that survives a sceptical CFO. |
| ROI | (Gross profit − all marketing cost) ÷ all marketing cost | “Did we make money?” | The number that decides whether to continue. |
Blended ROAS deserves a special mention for Singapore SMEs, because it solves a problem that costs local businesses real money: platform numbers double-count. Meta claims a sale, Google claims the same sale, and a summed figure describes a business twice its actual size. Blended ROAS has one revenue figure from your own system and one spend figure from your bank — it cannot double-count, and it needs no attribution theory to compute. It is the fastest honest number available to a small business.
The incrementality question, and what the evidence actually says
Every reported ROAS silently assumes the sale would not have happened otherwise. Often it would have. Retargeting is the clearest case: it advertises to people who already visited your site, many of whom were coming back regardless, and it therefore produces the best-looking ROAS in almost every account while contributing the least new revenue.
The most substantial public evidence comes from experiment providers running geographic holdout tests. Haus published an analysis of 640 Meta incrementality experiments and reported that Meta drove roughly a 19% average lift to brands’ primary KPI. Interestingly, their finding cuts against the usual narrative in one direction and with it in another: for direct-to-consumer measurement on a 7-day click basis, Meta tended to under-report its own contribution, while roughly 32% of the channel’s impact for omnichannel brands landed outside the DTC store entirely.
So the honest position is not “platforms exaggerate”. It is that a platform can overstate its claim on a specific conversion while understating its effect on the total business — and only a test can tell you which is happening to you.
The SME version of that test costs nothing but nerve: pause one channel for two to four weeks during a stable trading period, hold everything else constant, and watch total enquiries or total revenue rather than attributed figures. It is confounded by seasonality and is not a controlled experiment, so run it in a quiet month and repeat it before acting on the result. We go further into this in measuring social media ROI.
Five mistakes that turn a good number into a bad decision
- Comparing ROAS across channels as though it were like-for-like. Search captures existing demand and will nearly always out-ROAS social, which creates demand. Cutting social because its ROAS is lower often reduces the searches that search then harvests. Meta versus Google Ads covers this trade-off properly.
- Summing platform revenue. Both platforms claim the same order. Use blended ROAS instead.
- Judging a long sales cycle on a 30-day window. If your customers take three months to decide, a monthly ROAS reads as a failure for the entire ramp period. Match the measurement window to the buying cycle, especially in B2B — see B2B digital marketing in Singapore.
- Optimising to ROAS with discounting. Deep discounts lift revenue and ROAS while destroying margin. This is the most common way a business hits its ROAS target and its worst profit month simultaneously.
- Forgetting repeat purchase. First-order ROAS understates a subscription or repeat-purchase business, sometimes badly. If you know your repeat rate, judge acquisition on the customer’s expected contribution, not the first basket — while being honest that lifetime value is a forecast, not a measurement.
What to actually put on the wall
For most Singapore SMEs, three numbers are enough:
- Your break-even line — break-even ROAS for products, or maximum cost per qualified lead for services. Written down, derived from your own margin, reviewed when your costs change.
- Blended ROAS or MER, monthly — total revenue over total marketing cost, from your own systems. Immune to platform double-counting and to attribution changes.
- One incrementality test per quarter — even a scrappy on-off test. It is the only thing that tells you whether the first two numbers describe cause or coincidence.
Platform ROAS still matters — it is genuinely the best tool for deciding which ad set to scale on a Tuesday morning. It just should never be the number a business decision rests on.
Not sure whether your marketing is actually profitable? That is the question our performance marketing work is built around. See the results we have published, or talk to us about what your current numbers are leaving out.
Where to go next
- Performance marketing in Singapore — the hub for this cluster.
- How to read your monthly marketing report — where these numbers should appear, and in what order.
- Google Ads cost in Singapore and social media management cost — the cost side of the ratio.
- Is SEO worth it in Singapore? — the same profitability question applied to a channel with no media spend.
Frequently asked questions
What is the difference between ROAS and ROI?
ROAS is attributed revenue divided by ad spend — a media efficiency ratio that ignores every cost except the media itself. ROI is gross profit minus total marketing cost, divided by total marketing cost — a profitability measure. A campaign can have an excellent ROAS and a negative ROI, which is exactly what happens on thin margins.
What is a good ROAS in Singapore?
There is no universal figure. Your break-even ROAS is one divided by your gross margin: 3.3x at a 30% margin, 2.0x at 50%, 1.7x at 60%. Add roughly a quarter on top to cover agency fees, creative, tools, returns and delivery. Any published “good ROAS” that does not ask about your margin is describing someone else’s business.
Does GST affect my ROAS calculation in Singapore?
Yes, in two places. Revenue passed to the pixel usually includes 9% GST, so your true ROAS is about 8.3% lower than reported unless you strip it out. On the cost side, Google Ads charges 9% GST on Singapore-billed accounts, recoverable as input tax if you are GST-registered but a real cost if you are not, while GST-registered overseas suppliers should not charge GST once you supply your GST registration number.
How do I measure ROAS for a service business with no online sales?
You do not — you work backwards to a cost-per-lead ceiling instead. Multiply gross profit per job by your close rate and your lead qualification rate to get the gross profit value of one raw lead. That is your break-even cost per lead. Target a fraction of it so the marketing contributes profit rather than merely covering itself.
What is blended ROAS, and should I use it?
Blended ROAS is total revenue divided by total paid media spend, taken from your own systems rather than the platforms. It is worth using because it cannot double-count: Meta and Google both claim the same sale, so summing their reported revenue describes a business larger than yours. Blended ROAS needs no attribution theory and takes one spreadsheet.
What is incremental ROAS?
It is the additional revenue actually caused by the advertising, divided by the spend — measured with a holdout or geographic test rather than inferred from attribution. It matters most for retargeting and brand campaigns, which typically show the best reported ROAS and the weakest incremental effect. An analysis of 640 Meta experiments by Haus found an average lift of roughly 19% to brands’ primary KPI, with a meaningful share of the effect landing outside the online store for omnichannel businesses.
Related measurement guides
- Attribution models explained: which channel actually made the sale
- How to set up GA4 for a Singapore business
