Try this with last month’s numbers. Take the sales that Google says it drove. Add the sales Meta says it drove. Add whatever else you run. Then compare the total with what the business actually sold.
More often than you’d hope, the platforms have claimed more sales than exist.
Nobody is lying. Each platform counts a sale if its ad was somewhere in the customer’s path, by its own rules, inside its own window. A customer who saw a social ad, clicked a search ad and then opened an email is one sale to you and one sale each to three different reports.
Why this matters more than it used to
It used to be possible to stitch most of a customer’s path together. Browsers, privacy rules and consent choices have made that much harder, and the platforms now fill the gaps with modelled estimates. Those estimates are reasonable. They are also produced by the company whose product is being judged.
So the reported return on ad spend is partly measured and partly a platform’s educated guess about its own value. I’m not saying ignore it. I am saying it shouldn’t be the only evidence behind a six-figure budget decision, and very often it is.
The question that actually matters
The useful question isn’t “which ad got the credit?” It is “what would have happened if we hadn’t run it?”
That difference is everything. An ad shown to someone who was already typing your name into a search bar will look brilliant in a report and add almost nothing. An ad that introduces you to someone new might get no credit at all, because they bought three weeks later through another route.
Channels that sit at the end of the path get over-credited. Channels that start it get under-credited. Follow the reports faithfully and you will slowly move your money toward harvesting demand and away from creating it. Then, a year on, there is less demand to harvest and nobody can see why.
Better evidence, in order of effort
- Blended numbers. Total marketing spend against total revenue or total new customers. Crude, but no platform can inflate it.
- Switch-off tests. Pause a channel in one region or for a set period and see what happens to total sales. This is the most direct answer you can get, and most businesses have never done it once.
- Brand versus non-brand. Report them separately everywhere. It removes the biggest single distortion.
- Marketing mix modelling. A statistical model of sales against spend, seasonality, price and everything else, built on your own data. It used to be for very large advertisers. It is now within reach of mid-sized ones.
None of these is perfect either. But they fail in different ways from the platform reports, and that is what makes them valuable. When two independent methods agree, you can act. When they disagree, you have found something worth investigating.
What I’d do
Keep the platform reports for day-to-day management. That is what they are good at. For the bigger question of where next year’s budget goes, use at least one measure that the platforms don’t control, and run one honest switch-off test before you increase anything.
It is a strange industry habit to let the seller tell you how well the product worked. You wouldn’t accept it anywhere else.