Add up the conversions Google Ads, Meta and Google Analytics report for last month and you get more sales than you actually made. Each report is right by its own rules, and the rules differ. This guide explains why the numbers disagree, why the gap is growing as tracking disappears, and how to get one independent view of what actually drives sales.
Google Analytics, Google Ads and Meta each count conversions by their own definition: their own attribution window, clicks only or views as well, and only the touchpoints they can see. A buyer who touched two platforms is claimed by both. None of them sees TV, radio, out-of-home or print, and all of them see a shrinking share of users as consent rules and browser restrictions remove tracking. For an independent view across channels, use a marketing mix model. It works on aggregated spend and sales, needs no cookies or user-level tracking, and separates what marketing caused from what would have sold anyway. Keep the platforms' own tracking for what it is best at: optimising inside each platform.
Four mechanics explain most of the gap.
None of this is a bug. Each platform measures what it can see, with the rules it has chosen, and each is optimising for its own ads. The result is a set of numbers that are useful inside each platform and impossible to add up across them.
All of these systems depend on following a user from ad to purchase. That is getting harder. Consent banners under GDPR mean a large share of European visitors are never tracked at all. Apple's App Tracking Transparency cut what Meta can see on iPhones. Safari and Firefox limit cookies, and ad blockers remove pixels. The platforms fill the holes with modelled conversions, which are estimates the platform makes about conversions it did not observe.
The practical effect for a marketing team is that the reports agree less with each other every year, and less with the sales ledger. A measurement approach that needs to track individuals is measuring a shrinking sample and guessing the rest.
The two methods are often presented as rivals. They are better understood as tools for different jobs.
| Attribution (platform tracking, GA4, multi-touch) | Marketing mix modelling | |
|---|---|---|
| Unit of analysis | An individual user's path to one conversion | Weekly spend and sales for the whole business |
| Needs cookies or user tracking | Yes | No |
| Sees offline media | No | Yes: TV, radio, out-of-home, print, sponsorships |
| Separates incremental sales from the baseline | No: credits every tracked conversion to a touchpoint | Yes: estimates what would have sold with no marketing |
| Who runs it | The platform selling you the ads, or a tool built on their data | A model with no stake in any channel |
| Best used for | Bids, audiences and creatives inside a platform, day to day | How much to spend in total and how to split it across channels |
A marketing mix model is a statistical model of how sales respond to spend per channel. It is built on aggregated history, controls for seasonality, price changes and other large non-marketing effects, and estimates the baseline: the sales that would have happened with no marketing at all. What is left is the incremental effect of each channel. Because it never follows a user, it is unaffected by consent rates, cookies or attribution windows, and it treats a TV spot and a search campaign the same way.
An example of what the model shows
Platform reports typically show branded search with a very high return and TV with none, because TV has no pixel. A marketing mix model usually shows some of the branded-search effect belongs to the channels that created the demand, TV among them, and that a share of the search conversions would have happened anyway. The exact split is the model's view for that business, with a range, not a general rule.
The platforms' own tracking is the best thing there is for optimising inside each platform. Google and Meta see far more signal about their own auctions, audiences and creatives than any outside model can, and their optimisers act on it in real time. For deciding which ad, audience or keyword to run, and for bidding, keep using their tracking.
The limit is that platform numbers are not a currency you can use across channels. The budget questions, how much in total and how to split it, sit across channels, so they need a measure that works across channels. In a well-run setup the model sets the budget level per channel and the platforms execute it. Platform data is not wasted in the process: together with what your team already knows about each channel, it informs the model's starting assumptions, and the model then measures each channel against your actual sales.
There are three routes to a marketing mix model, and the right one depends on the team you have.
Whichever route you take, three things decide whether the model is worth trusting: whether it includes every channel you spend on, offline included; whether its starting assumptions are set with care and shown to you; and whether it is verified against what actually happened once the forecast period has passed. For the questions to ask, see how to choose a marketing mix modelling vendor.
Because they count by different rules. Meta counts view-through conversions and uses its own window; Google Analytics only sees sessions that arrived from a click and applies its own attribution model. Each sees only its own touchpoints, so a buyer who touched both is claimed by both.
Each is right by its own definition and wrong as a total. The number to plan from is the one in your sales ledger. A marketing mix model explains that total by channel, including the sales that would have happened without any marketing.
With a method that does not need to track individuals. Marketing mix modelling works on weekly spend and sales for the whole business, so consent rates, cookie restrictions and attribution windows do not affect it. Experiments, such as geo tests or varying one channel's spend over time, are the other tracking-free method and can be fed into the model.
Multi-touch attribution follows tracked users and splits credit for each conversion across the digital touchpoints before it. Marketing mix modelling estimates how total sales respond to spend per channel, offline included, and separates incremental sales from the baseline. The first is for running campaigns inside digital channels; the second is for setting the budget.
No. It is the best tool for optimising inside each platform. Use it for bids, audiences and creatives, and use the model for the total budget and the split across channels.
No. Analytics remains the tool for what happens on your site. The model answers a different question: which spend caused which sales, across every channel, and how much would have sold anyway.
If your sale happens weeks after the click, or through an adviser, read marketing mix modelling for banks, fintech and long sales cycles. For how the model is built, read our complete guide to marketing mix modelling. For how platform data and experiments go into one model, read triangulation in marketing measurement. To see what the model shows for your channels, book a demo.