Picture a Google Ads account showing a ROAS of 6 in the dashboard: for every euro spent on advertising, six euros in tracked revenue come back. Textbook-perfect. The agency runs the monthly call, the slides are green across the board, the marketing director is happy.
Same period, same board meeting. The finance team presents the year's revenue: +8% vs the prior year. Good, sure. Except that ad spend grew by 40%.
Both curves can't be true at the same time.
This is the most common pattern I see in Google Ads and Meta account audits. It's not rare. It's not an edge case. It's the norm for Italian SMEs that report a high platform ROAS. The dashboard tells a story of efficiency; the P&L tells a story of slow growth at rising cost. They look like two different companies — but it's the same one.
The reason is that the two numbers are measuring different things. And almost nobody inside Italian companies is in a position to see that.
There are at least three numbers we all call "ROAS" that are actually three different things. Knowing the difference is the prerequisite for any serious conversation about performance.
Platform ROAS
This is the number you see when you open Google Ads or Meta. It calculates: attributed revenue from the platform divided by spend on the platform. It's the highest of the three, and the most optimistic — because every platform attributes conversions using its own model, almost always last-click tuned in its own favor. If Google and Meta both report a ROAS of 5, and you spent €10,000 on each, the two platforms together are telling you they generated €100,000 in revenue. Your ERP, meanwhile, recorded €70,000.
Real ROAS
This is the ROAS calculated net of traffic that was already yours. People searching for your brand on Google. People already on your email list who click a retargeting ad. People who would have found your site anyway because they already knew you. That audience converts at low cost and produces sky-high apparent ROAS — but the campaign isn't bringing them in. It's intercepting them after the fact. Once you remove that layer, ROAS falls. Often substantially.
Incremental revenue
This is the most important figure and the hardest to calculate precisely: the additional revenue you have today that you wouldn't have without the campaign. It's the delta between a world with the campaign and a counterfactual world without it. It's estimated using specific methods (geo-tests, holdout tests, marketing mix models) and almost no SME has these tools at hand. But even a rough approximation beats any platform ROAS.
Platform ROAS is what the platforms tell you. Real ROAS is what pays the bills.
Platform ROAS serves Google and Meta to justify their ad spend. Real ROAS serves you to understand whether you're actually growing or paying to be visible to people who would have bought from you anyway.
These are two different jobs. Mixing them up costs tens of thousands of euros per year — and the problem is that the first is displayed in your dashboard every morning, while the second one you have to build yourself.
The gap between these three figures isn't random. It has three structural causes, and until you recognise them you'll keep misreading data even with the most sophisticated tools.
Cause one: branded keywords
This is the most widespread problem and I've described it in detail in another piece: in 7 out of 10 Italian accounts there's budget allocated to keywords that include the brand name. People searching for your name already know you. That click would have converted for free from the organic result. The platform takes the credit, ROAS goes up — but the incremental revenue is zero, because that money would have arrived anyway. All a branded campaign does is replace the organic click with a paid one.
Cause two: last-click attribution (and its close relatives)
When a customer enters your funnel, the average journey isn't "sees ad → buys". It's: sees an organic Instagram post, then a friend's story, then searches on Google, then clicks a Meta retargeting ad, then comes back organic, then finally buys. Six touchpoints. At least. But the default attribution model on most platforms tends to give almost all the credit to the last measurable interaction, or to a handful of tracked touches. Everything in the pre-funnel — which actually built the purchase intent — gets erased from the calculation. Result: the campaign that closes (usually the lowest-funnel one, like retargeting or brand search) looks like it's working miracles. The one that opens (prospecting, awareness) looks useless. You cut it, and a few months later the funnel empties.
Cause three: double-counting across platforms
The same conversion gets claimed by both Google and Meta at the same time. A person sees an Instagram ad, a few days later searches for the product on Google, clicks a Search ad, and buys. Meta claims the conversion (it showed the ad first within the view-through window). Google claims the conversion (it had the last click). If you add the two platform ROAS numbers together, you get a figure larger than actual revenue. Often 20–40% larger. Neither one is lying — they're using different attribution logic and talking past each other.
Three independent causes, acting together, reinforcing each other. That's why the average platform ROAS in the Italian market is structurally optimistic — not because anyone is cheating, but because the measurement system is built that way.
There are rigorous methods for measuring real ROAS: incrementality tests, geo-tests, marketing mix models. They're powerful but require time, clean historical data, and someone who knows how to interpret them. For most Italian SMEs, that level of sophistication makes sense to build later, not now. What does make sense to do immediately is a quick gap estimate — three steps you can start next week without buying anything.
Step one: separate branded from non-branded
In Google Ads, create a segment or filter that isolates campaigns (or ad groups) containing your brand name in the keywords. In GA4, do the same at session level: branded vs non-branded search terms. Calculate the ROAS for each segment separately. The branded one will be very high. The non-branded one is the baseline you should actually reason from — because that's where you measure how much you're paying to acquire customers who didn't already know you.
Step two: compare the delta in spend with the delta in revenue
Take 12 months of ad spend and 12 months of revenue. Calculate the year-on-year change for both. If spend grows 40% and revenue grows 8%, the revenue per incremental advertising euro is embarrassingly low — the campaign is producing far less than the dashboard suggests. If spend grows 20% and revenue grows 30%, it's the opposite: there's real leverage. This is a rough approximation — it doesn't account for seasonality, product launches, or market dynamics — but it's already a more honest sanity check than any platform ROAS.
Step three: estimate a marginal ROAS
Take two comparable periods (for example Q1 this year vs Q1 last year). Calculate: (Q1 revenue this year − Q1 revenue last year) / (Q1 ad spend this year − Q1 ad spend last year). This is the marginal revenue per marginal euro: a very rough but useful proxy for incremental ROAS. If it's 1, every extra euro in advertising brings one extra euro in revenue — below break-even on margins for almost every sector. If it's 3, there's real leverage. If it's negative, the campaign is consuming without producing.
When a team starts reasoning about real ROAS instead of platform ROAS, three things typically happen — usually in this order.
The first is that the campaigns deemed winning change. Brand search and retargeting campaigns stop being the star performers — not because they should be switched off, but because it becomes clear they were "harvesting" more than "generating". Prospecting, awareness, and cold-audience campaigns stop being the black sheep — because it becomes clear that's where the pipeline was being built that the others were then closing.
The second is that the conversation with leadership changes. You stop having to defend "ROAS 6 in the dashboard" in front of a board that sees slow growth. You start reasoning in terms of customers acquired, real acquisition cost, lifetime value, incremental revenue. These are smaller numbers, less dramatic — but they're numbers the CFO understands. And ones the board can compare with other investment lines.
The third is that the relationship with the agency changes. It becomes much harder for whoever manages your campaigns to sell themselves on platform numbers. It becomes much easier for both of you to focus on what matters. This transition isn't always painless — some agencies welcome it, others resist it — but the result six months in is almost always a healthier and more productive budget allocation.
Real ROAS isn't calculated — it's built. And it's built upstream, when you plan, not downstream when you measure.
There's an entire industry of attribution software that promises to fix everything described above. These tools are useful — some are excellent. But they tell half the truth: none of them can repair ex-post an architecture that was wrong ex-ante.
If you designed your campaigns without clearly separating acquisition audiences from retention audiences — and in most Italian accounts that separation simply doesn't exist — no attribution tool will reliably tell you how much one produces versus the other. It will give you numbers, certainly. But they'll be numbers measuring a jumble instead of two distinct things.
The problem, in other words, isn't measurement. It's planning. A genuinely high real ROAS isn't found by looking at data more carefully: it's built by deciding, upfront, what share of budget goes to bringing in new customers, what share goes to getting existing customers to buy more, and what small portion goes to defending the brand. Three different objectives, three different KPIs, three different expected ROAS.
When that architecture is in place, any measurement tool — even just bare Google Ads and GA4 — starts giving you consistent answers. When it isn't, even the best attribution model in the world will tell you things you can't act on.
This is why, every time a new client asks us to "optimise their campaigns", we don't touch the campaigns in the first week. We look at how they're planned. And almost always, that's where the real work needs to happen.
Observations based on Google Ads and Meta Ads audits conducted on Italian accounts across 18 sectors, 2022–2025. ROAS gap estimates are based on recurring patterns observed in audits, not a formal sample study. To discuss an audit of your account or set up a planning structure that separates acquisition from retention, write to commerciale@goodea.it.