Why ROAS can mislead you: how to measure the real impact of Google and Meta campaigns

Google Ads says ROAS 6, Meta says ROAS 5, the management system says revenue has grown little or not at all. If you add up the sales the two platforms attribute to themselves, you exceed the company’s actual sales. Neither report is lying: each counts conversions by its own rules, which tend to favour the platform itself.

The short answer is this: the ROAS you read in the platforms measures the revenue the platform attributes to itself, which is different from the revenue the campaigns caused. To know how much Google and Meta really deliver you need three more numbers alongside ROAS: margin (that is POAS or break-even ROAS), the share of new customers and a check on the total, MER. When the doubt is serious, you run an incrementality test. In B2B, where the sale closes in the CRM weeks after the click, the benchmark is the value campaigns bring into the pipeline.

In the campaigns I manage I always start from cost per lead: I look at ROAS, but as an indicator to be checked.

Why doesn’t platform ROAS match actual sales?

Because each platform attributes the conversion to itself according to its own time windows and models, without knowing what the user did on the other channels. The causes are almost always five.

Attribution windows

In Google Ads the default conversion window for clicks is 30 days (adjustable from 1 to 90), with 3 days for engaged video views and 1 day for view-throughs (Google Ads guide to conversion windows). On Meta the most common setting is 7 days after the click and 1 day after the view, among the windows still available from 12 January 2026 (Meta for Developers). If someone clicks, thinks about it for three weeks and then comes back by typing the website address, for Google Ads that sale still belongs to the campaign, even if in the meantime the salesperson closed it on the phone.

View-through conversions

A view-through is a sale attributed to someone who saw the ad without clicking on it. In remarketing it rewards the ad just because it appeared in the feed of someone who would have bought anyway. Meta has already narrowed this logic: from 12 January 2026 it removed the 7-day and 28-day view windows from the API, and since the end of March 2026, for campaigns optimised for conversions on the website or in store, only a click on the link counts as a click-through. Likes, shares, saves and videos watched for at least 5 seconds end up in a separate category, engage-through (Meta’s announcement of 3 March 2026). Meta ROAS in 2026 and in 2025 are measured with two different yardsticks.

Double counting between Google and Meta

A user sees a video on Instagram, two days later searches for the company name on Google, clicks the ad and buys. Meta counts the sale (view or click within the window), and so does Google (click on the search ad). There’s one lot of revenue; in the reports it becomes two.

Brand search claiming sales already made

Campaigns on your own brand almost always have the highest ROAS in the account, because they catch people who are already looking for you, and some of those clicks would have come through organic anyway. They make sense for protecting the results page from competitors, but they need to be read separately: mixed in with acquisition they push up the average ROAS and hide the campaigns that are struggling. The same goes for Performance Max, which without brand exclusions tends to spend precisely where conversion is easiest; we covered this in our guide to Performance Max campaigns.

Incomplete tracking

The last problem goes in the opposite direction: without cookie consent some conversions aren’t recorded and the platforms estimate them. With a badly configured Consent Mode, ROAS comes out too low and Smart Bidding learns from the wrong data; we explain how to fix it in Enhanced Conversions and Consent Mode.

Attributed revenue or incremental revenue: what’s the difference?

Attributed revenue is what the platform links to the campaign according to its rules. Incremental revenue is what wouldn’t have existed without the campaign. Only the second tells you whether the spend produced anything, and it’s measured with an experiment, comparing a group exposed to the ads with one that doesn’t see them.

Google Ads. Lift studies are found in the Experiments tab. User-based Conversion Lift splits the audience into a group that sees the ads and a control group that doesn’t; according to Google’s documentation it requires at least 1,000 observed conversions, a minimum budget of 5,000 dollars per campaign and a recommended duration of at least 14 days. It works on Search, Display, Video, Demand Gen, App and Performance Max. There’s also a geography-based variant, which compares areas where the campaign is running with areas where it’s switched off.

Meta. Conversion Lift tests split the audience into a group that sees the ads and a control group that doesn’t; according to Meta’s documentation access is limited and must be requested from your Meta representative. There’s also incremental attribution, a setting that estimates which conversions were caused by the ads and not just which ones happened after a click or a view.

Home-made geo tests. If volumes aren’t enough for the platforms’ tests, you can switch off (or double) a campaign in some regions while leaving it unchanged in others of similar size, and compare actual sales from the management system or CRM over four to six weeks. It’s less precise than a lift study, but it uses your own sales data.

Holdouts. You exclude part of the audience (for example 10% of the customer list) from a remarketing campaign and measure whether it buys less than the others. If it buys the same, you’re paying for conversions you would have had for free.

An incrementality test makes sense when an item of spend is large and the doubt is real: the remarketing that “converts brilliantly”, the brand campaign, Meta claiming half of all sales.

How do you calculate break-even ROAS?

Break-even ROAS is calculated by dividing 1 by the gross margin percentage: with a 40% margin, break-even is 1 / 0.40 = 2.5. Below that value every euro spent on advertising costs you more than it earns you, even if the ROAS looks “good”.

ROAS works on revenue; the business lives on margin. A hypothetical example with two product lines and 1,000 euros of spend each:

	Line A	Line B
Attributed revenue	€5,000	€3,000
ROAS	5	3
Gross margin	20%	50%
Gross profit	€1,000	€1,500
POAS (profit / spend)	1.0	1.5
Break-even ROAS (1 / margin)	5.0	2.0

Line A has the higher ROAS and is breaking even: the margin covers the advertising and nothing more, even before logistics and returns. Line B delivers more with a lower ROAS. Optimise on ROAS and you move budget towards the wrong product.

POAS (profit on ad spend) is the gross profit generated divided by ad spend. Above 1 the campaign pays for itself and contributes to margin; below 1 it eats into margin. To use it in Smart Bidding you pass Google Ads and Meta the margin instead of the price as the conversion value: a feed and tracking job you set up once.

Break-even ROAS is also used to set the target: a tROAS set below break-even asks the algorithm to make you lose money efficiently.

New customers or customers who would have come back anyway?

A sale to a long-standing customer says little about the impact of advertising; a sale to a new customer says a lot. ROAS counts them the same way, so a campaign that lives on repeat purchases can have splendid numbers without growing the customer base.

Google Ads lets you manage this difference with the new customer acquisition goal, available on Search, Performance Max, Shopping and Demand Gen. There are three modes: bid higher for new customers (adding extra value to the conversion), also distinguish high-value new customers, or bid for new customers only. To recognise existing customers it uses uploaded customer lists and data from the website tag.

In lead generation, the “new customers only” mode avoids paying for the click of a customer who searches for your name to find the customer service number. On Meta the first step is to exclude customers and leads already in the CRM from acquisition campaigns. In reports, cost per new customer compared with the average first-year value says much more than aggregate ROAS.

How do you measure campaigns in B2B and lead generation?

In B2B platform ROAS barely exists, because the sale happens in the CRM weeks or months after contact. The useful metric is the value each campaign brings into the pipeline: how many qualified opportunities, for what amount, and how many become contracts.

Take an engineering company running campaigns on Google and Meta, with figures made up for illustration. Google brings leads at 120 euros each, Meta at 45 euros. Looking only at cost per lead, Meta wins easily. Then you open the CRM: of the Google leads, one in three becomes an opportunity; of the Meta leads, one in fifteen. Cost per opportunity is 360 euros for Google and 675 for Meta. The picture is reversed, and you only see it if conversions make it all the way to the CRM.

To get there you need three things.

1. An identifier that follows the lead. GCLID for Google, Meta’s lead ID for instant forms, or hashed email and phone. Without it, the CRM doesn’t know which campaign the contact came from.
2. CRM stages sent back to the platforms. On Google this is done with enhanced conversions for leads, which Google recommends instead of the old offline conversion import and which connect through Data Manager. On Meta there’s the Conversions API for CRM integration with the Conversion leads goal, which optimises towards contacts that become qualified rather than towards people who fill in the form. Meta requires at least 200 leads a month from instant forms and daily uploads.
3. A value for each stage. Qualified lead, opportunity, contract: each with an estimated financial value. That way Smart Bidding learns that a 50,000-euro opportunity is worth more than a form filled in by a student.

How to build the flow from the form all the way to the invoice is something we described in from lead to contract: connecting CRM and revenue.

What is MER and why is it worth looking at?

MER (marketing efficiency ratio), also called blended ROAS, is the company’s total revenue in the period divided by total ad spend across all channels. It doesn’t depend on any attribution model, so it doesn’t suffer from double counting.

MER doesn’t tell you which campaign works; it tells you whether the system as a whole holds up. If platform ROAS rises and MER stays flat, the platforms are fighting over the same sales. If you increase spend by 30% and MER collapses, the extra money hasn’t found new demand.

For B2B the useful version is a pipeline MER: value of new opportunities created in the month divided by total ad spend. To read it you need spend and results from all channels in one place: for the advertising side we use RunAlytics, the dashboard that brings together data from Meta, Google Ads, TikTok and LinkedIn campaigns, while revenue comes from the management system or CRM.

Which metric to use and when

No metric is enough on its own. This is the map we use depending on the question to be answered.

Metric	What it tells you	When to use it
Platform ROAS	Revenue Google or Meta attribute to themselves per euro spent	Day-to-day optimisation within a single platform, comparing campaigns on the same channel
Break-even ROAS	The minimum ROAS below which the campaign loses margin	To set the tROAS target and judge whether a ROAS is good
POAS	Gross profit generated per euro spent	E-commerce with different margins across products or categories
Cost per new customer	How much it costs to acquire a customer you’ve never seen before	When the goal is to grow the customer base
Cost per lead	How much a lead costs	Lead generation, in the first few weeks and for quick comparisons
Pipeline value	Opportunities and amounts generated by each campaign, read from the CRM	B2B with a long sales cycle
MER / blended ROAS	Total revenue divided by total spend	Monthly check on overall health, budget decisions between channels
Incremental lift	Sales that wouldn’t have happened without the campaign	Before confirming or cutting a major item of spend

For a broader view of how to bring together costs, revenue and overall return, there’s our guide to digital marketing ROI. If the numbers from the platforms, Analytics and the CRM don’t match, before choosing a metric it’s worth checking the data from click to sale.

Checklist: how to put measurement back in order

• Separate brand campaigns from acquisition campaigns in your reports, on Google and on Performance Max.
• Check the conversion windows of each action in Google Ads and the attribution setting on Meta, and note the dates of Meta’s January and March 2026 changes before comparing with last year.
• Calculate break-even ROAS for each product or service line and compare it with the targets you’ve set.
• Add up the sales attributed by all the platforms and compare them with actual revenue for the period.
• Upload an up-to-date customer list and consider the new customer acquisition goal.
• Connect the CRM to the platforms with enhanced conversions for leads and the Conversions API, with a value for each stage.
• Choose the item of spend you have most doubts about and plan an incrementality test or geo test.

Mistakes to avoid when reading ROAS

Adding up the platforms’ ROAS. The total almost always exceeds actual revenue. If you want a single number, use MER.

Comparing ROAS with industry averages. A ROAS of 3 is excellent with 60% margins and disastrous with 15% margins. Your benchmark is your break-even point.

Cutting prospecting campaigns because their ROAS is low. Top-of-funnel campaigns feed remarketing and brand search. Switch them off and after a few weeks the ROAS of the campaigns that seemed the best falls too.

Optimising B2B on form fills. The algorithm finds people who fill in forms, who often aren’t the ones who sign contracts. Without feedback from the CRM, you lower cost per lead and worsen quality.

Trusting view-through in remarketing. A 10% holdout makes clear within a few weeks how much of that ROAS is down to the ad.

What changes for lead generation, CRM and performance

The most concrete change concerns where the budget is decided. With platform ROAS, decisions are made inside Google Ads and Meta, by their rules. With pipeline value or margin, they go through the CRM and the management system, that is through data the company controls.

For anyone doing lead generation this means judging marketing on opportunities rather than leads, and therefore agreeing with sales on what a qualified lead is and how it’s marked in the CRM. Bidding algorithms receive a better signal and, after the learning phase, bring fewer, better-suited leads. And the comparison between Google, Meta and LinkedIn is made on the same number, read in the same place.

The foundation is clean tracking, well-managed consent, up-to-date CRM stages and an automatic connection to the platforms. It’s the least visible part of our analytics and data service, and it’s what makes everything else reliable.

Want to know how much your Google and Meta campaigns really deliver?

We carry out an audit of campaigns and tracking: we check how conversions, attribution windows and Consent Mode are set up, how much brand and remarketing weigh on the ROAS you see, whether the CRM sends data back to the platforms and where there’s spend that isn’t bringing in new customers. At the end we tell you what to fix first and whether an incrementality test makes sense for your volumes.

You can book a call with Federico or write to us via the contact form. You’ll find how we manage campaigns on the pages about Google Ads and Meta ads. If your measurement is already in order, we’ll tell you.

Frequently asked questions

Can Google Ads ROAS and Meta ROAS be added together?

No. The two platforms can both attribute the same sale to themselves, so the sum almost always exceeds actual revenue. For an overall view use MER, that is total revenue divided by total ad spend.

What’s a good ROAS?

It depends on the margin. The minimum value is break-even ROAS, 1 divided by gross margin: with a 25% margin you need a ROAS of at least 4 just to cover the advertising.

How do you measure campaign incrementality?

With an experiment that compares a group exposed to the ads with a control group: Conversion Lift in Google Ads and Meta, geo tests across different regions, or a holdout on part of the audience. For user-based lift, Google requires at least 1,000 conversions and a budget of 5,000 dollars per campaign.

Does ROAS make sense in B2B?

Not much, because the sale closes in the CRM after weeks or months. It’s better to measure cost per opportunity and pipeline value, sending CRM stages back to Google and Meta with enhanced conversions for leads and the Conversions API.

Why did my Meta conversions drop in 2026 when I didn’t change anything?

From 12 January 2026 Meta removed the 7-day and 28-day view attribution windows from the API, and since the end of March 2026 it only counts a click on the link as a click-through, moving other interactions into engage-through. The drop in your reports may be down to this rather than to the campaigns.