How do I know which ad channel actually made money?
Why the dashboards disagree
Each platform counts a conversion it believes it influenced, using its own attribution window and its own modelling, so a customer who saw a Meta ad on Monday, searched your brand on Google on Wednesday and finally bought on Thursday shows up in both accounts at once as a completed sale. Neither platform is lying. They are answering a narrower question than the one you asked.
Brand search makes it worse. Google Ads takes credit for capturing demand that Meta or a creator created, so it usually reports flattering numbers while adding relatively little incremental revenue.
Add iOS signal loss and modelled conversions and the picture blurs further. Sum the platform-claimed revenue for a month, then look at your actual store revenue. When the first number is far larger than the second, you know exactly how much to trust channel ROAS.
Three ways to get closer to the truth
Blended metrics first, and use them for decisions. Total marketing spend divided by total new customers gives you blended customer acquisition cost, and total revenue divided by total spend gives you blended ROAS. Neither number can be inflated by attribution games, because both are drawn straight from your own bank of truth.
Next, a post-purchase survey on the thank-you page. One question: where did you first hear about us? Open text or a short list. It is self-reported and imperfect, but it reliably surfaces channels the pixels miss entirely, particularly WhatsApp forwards, word of mouth and creator mentions.
Then holdout tests, which are the only real measurement here. Switch a channel off for two weeks, or geo-split by region, and watch total revenue. If pausing a channel with a reported 4x does nothing to your overall sales, it was harvesting demand rather than creating it.
Making it a routine
Review blended CAC and contribution margin weekly. Read survey responses monthly for shifts. Run one holdout test a quarter on your largest or most suspicious channel, and plan it properly, because a badly timed test during a festival sale tells you nothing at all.
Then use platform ROAS for what it is genuinely good at, which is comparing creatives and audiences inside one account. It was never designed to arbitrate between channels.
Related questions: Should I sell on marketplaces or my own website? · How do I lower my blended CAC? · Why is my Shopify store slow?.
Want this handled: E-commerce growth · All answers.
Add up what every platform claims and compare it against your actual Shopify revenue. The gap tells you how much double-counting is going on. Meta and Google both claim the same order, so channel-level ROAS in the dashboards is close to fiction. Blended numbers, a post-purchase survey and switching a channel off are the three tools that give real answers.
Generally not below a substantial monthly spend across several channels, because these models need volume and history to say anything reliable. Blended metrics plus quarterly holdout tests give most brands the same practical answers at far lower cost.
Neither as your only view. Last-click over-credits search and retargeting. First-click over-credits awareness. Use blended numbers for budget decisions, and keep the attribution views for diagnosing what is happening inside a channel.
Ready to replace guesswork with a growth engine?
Book a 30-minute strategy call. We’ll show you exactly where your funnel is leaking, before you spend a dollar.