TLDR: Your Google Ads ROAS figure looks clean. It probably isn’t. Last-click attribution, blended reporting, view-through inflation, cross-device gaps, and seasonal distortion all conspire to make your numbers look better or worse than they actually are. This blog breaks down each problem and what to do about it.
ROAS is the metric most Google Ads teams live and die by. It’s clean, it’s simple, and it gives you something to put in a slide. The problem is it’s often wrong, not because the maths is off, but because the data feeding it is incomplete.
Last-click attribution ignores 80% of the journey
Last-click attribution does exactly what it says: it gives 100% of the credit for a conversion to whatever the user clicked on last. If that final click was a Google Shopping ad, Shopping gets all the credit. If it was a direct visit, direct gets all the credit, regardless of everything that came before it.
The reality of how people actually buy something looks nothing like that. This is more than likely what happens, Someone needs new trainers. They see an influencer on social. They click a competitor’s ad. They read an organic blog about the brand. They search again, find a shopping ad, and convert. That whole journey gets credited to Google Shopping. But they came through organic, social, direct, and a competitor ad first.
A user might click a Google ad 20 times and an organic listing 10 times before finally converting via a direct visit. On last-click attribution, direct gets 100% of the credit. Every other touchpoint gets zero.
This creates a specific and expensive blind spot for upper-funnel activity. YouTube, for example, is a brand awareness channel. It almost never drives direct conversions, in the same way a TV ad rarely causes someone to immediately buy something. But TV advertising works. We know it works because of studies like MMM that measure the halo effect across channels. YouTube deserves the same scrutiny.
The fix: Move to data-driven attribution (DDA). This is what Google recommends, and for good reason. Instead of giving all the credit to the last click, DDA distributes credit across touchpoints based on their actual contribution. It is only as good as the tracking data you feed it, so getting your tracking right is non-negotiable, but it gives a far more honest picture of what is actually driving performance.
Blended ROAS hides channel-level losses
Blended ROAS looks at your overall ROAS across all campaigns, where strong performance quietly subsidises poor performance. And the most common culprit is brand vs non-brand.
Brand campaigns convert at a much higher rate than non-brand campaigns. That makes sense: someone searching directly for your brand name has higher intent. The problem is that a blended ROAS figure mixes brand performance in with everything else, so your generic campaigns, which typically have much higher volume but much lower efficiency, get obscured.
There is also a more uncomfortable question underneath brand bidding: are you paying to capture traffic that would have converted organically or directly anyway? Sometimes yes, sometimes no. With the rise of broad match, AI Max, and Performance Max, competitors are increasingly able to show above your organic listings on your own brand terms, which makes brand bidding harder to avoid. But the answer to that question is very different to ‘let’s roll it into the blended number and not look too closely.’
The fix: Segment brand and non-brand reporting separately. Look at campaign-level ROAS rather than account-level ROAS. Dig into product-level reporting within your non-brand campaigns to understand what is actually performing and what is dragging the average down. The blended number is useful for a top-line view, it is not useful for making budget decisions.
View-through conversions inflate the number
YouTube and Demand Gen present a specific measurement problem. Direct conversions, someone sees an ad, clicks it, and converts, tend to undervalue these channels, because that is not really how they work.
The result is that neither metric alone gives you an accurate read. You end up either dismissing a channel that is genuinely building demand or crediting it for conversions it did not really drive.
The gold standard for understanding what YouTube is actually doing is running uplift studies. A brand uplift study measures the increased likelihood of a user converting after seeing an ad, it works like a survey shown after ad exposure. The catch is that this requires significant investment: around £30,000 across four weeks is the ballpark. If you do not have that kind of spend going through YouTube, it may not be the right channel for you yet.
The more accessible alternative is a search uplift study. This is an A/B test that measures the increase in search volume across a keyword set, typically this could be branded terms. If users who saw your ad are 50% more likely to search for your brand or product, you can work backwards: apply your click-through rate, apply your conversion rate, and get to an estimated revenue figure and from there, a minimum ROAS. It is not a perfect number, but it is significantly more reliable than either view-through or direct conversions in isolation.
The fix: Exclude YouTube and Demand Gen from your standard ROAS calculation. Run search uplift studies to understand what these channels are actually contributing. Do not try to measure brand awareness channels the same way you measure lower-funnel conversion campaigns.
Cross-device journeys break the attribution chain
Most customer journeys do not happen on a single device. Someone might discover a product on Instagram on their phone, research it on a laptop, and convert on a tablet after seeing a retargeting ad. In a cookie dominant, last-click world, whichever device drove that final click gets the credit. The rest of the journey disappears.
This is where Enhanced Conversions in Google Ads become important. When cookie-based tracking breaks down, because someone has switched device, switched browser, or cleared their cookies. Enhanced Conversions uses hashed first-party data (name, email, phone number) to match up those touchpoints across sessions. If a user is logged in to the same Google account across devices, their journey can be stitched back together.
The practical example: a user logs into a website through their Google account on their phone, later searches on Chrome on a laptop still logged in to the same Google account, watches a YouTube ad on their TV through the same account, and then converts through a Google ad, putting in the same name, email, and address at checkout. Enhanced Conversions can connect those dots. Without it, much of that journey is invisible.
The fix: Set up Enhanced Conversions in Google Ads. It does not require you to do anything beyond the initial configuration, once it is live, it uses existing first-party data to match journeys that would otherwise be lost. Combine this with DDA, and your attribution picture becomes significantly more complete.
Seasonality distorts your baseline
A strong ROAS number in December does not mean your campaigns are suddenly performing better. It might just mean it is December. Seasonal demand peaks, like Black Friday, Christmas, summer heatwaves, inflate ROAS in ways that have nothing to do with campaign quality, and a business that mistakes seasonal uplift for structural improvement will make very bad decisions in January.
For example, some shoe brands will see a spike in demand during warm weather. That makes intuitive sense, people go outside, notice they need sandals, and buy. But the week after a heatwave tends to be noticeably weaker, because the demand that would have been spread across that period has already been captured. If you only look at the hot week, you overestimate what your campaigns are delivering. If you only look at the week after, you underestimate.
The same dynamic plays out with Black Friday. A brand that ran aggressive offers and captured outsized Q4 revenue needs to understand whether that came at the cost of Q1 to Q3, or whether it genuinely represented incremental growth. A great Q4 ROAS that masks three quarters of underperformance is not a success story.
The fix: Know your products and know your trends. Break reporting down to a granular enough level that seasonal patterns become visible rather than hidden in monthly or quarterly averages. Year-on-year comparisons only mean something if you understand why one year looked different to the last. Context is not optional, it is the only thing that makes the numbers useful.
What does good look like?
Good attribution is not one thing. It is a combination of the right attribution model (DDA rather than last-click), the right tracking infrastructure (Enhanced Conversions, solid GA4 setup), the right reporting structure (brand and non-brand separated, campaign-level visibility rather than just account-level), and the right analytical context (seasonal patterns understood, upper-funnel channels measured appropriately).
The teams that get this right are not the ones with the best ROAS number. They are the ones who know whether to trust the number and what to do when they do not.
That is exactly the kind of visibility tools like ASK BOSCO® is built to give you. It pulls all of your marketing and ecommerce data into a single platform, so instead of switching between GA4, Google Ads, and half a dozen other dashboards, you have one place where everything is connected and comparable. You can graph performance across any time period without hitting GA4’s 13-month limit. Segment by channel, and get the cross-channel view that blended platform reporting cannot give you.
And with the new ASK BOSCO® AI Studio, you don’t even need to build the report. Just ask the question and get a tailored report back instantly. Whether you want to understand what happened to ROAS across a specific campaign last month. Or how performance looked year on year around a peak trading period, you can get the answer without writing a single query or waiting on an analyst. If you want cleaner, more reliable visibility across your paid media performance. Get in touch with the Modo25 team to talk through your measurement setup.
