In June 2026, we started working with an online store in Romania that sells professional cosmetics. Its Facebook and Instagram ads had brought in 7 orders in total, worth about a third of what the ads cost (a return on ad spend, or ROAS, of 0.35x). Google Ads, meanwhile, showed a ROAS of 22x in some reports. Same website, same products. Those two numbers couldn't both be right, so before writing a single new ad, we checked what each account was actually counting.
One campaign was selling too well to be true.
A campaign from December 2024 showed 3,294 purchases from 4,904 clicks. That would mean two out of every three people who clicked the ad bought something. So we went looking for what was behind the number.
Page loads. The tracking code had been set up wrong, so just opening a page got recorded as a purchase, whether or not anyone bought anything. That's where the 22x came from.
Google was bidding for YouTube subscribers.
The account was set to count 17 different things as conversions, and half of them were duplicates. Two YouTube actions, including channel subscriptions, were set as primary, and the purchase wasn't. Google's automated bidding only chases primary actions, so it was hunting for people who'd subscribe, not people who'd buy.
Three other campaigns showed hundreds of conversions with a value of zero. They were counting something, just not sales. Without the order value, Google can't tell a $10 cleanser from a $90 set. The one campaign we could trust was the one showing the store's products in Google with a photo and a price, and it brought back about 1.2 times what it spent.
On Facebook, the ads were looking for people who add to cart.
About 78% of the Facebook sales budget was optimized for add-to-carts and checkouts, not purchases. Meta delivered exactly that: people who added something to the cart and left. Orders: 7.
The Meta pixel, the code that tells Facebook what people do on a website, had been removed from the store's site. It last sent data in March. From then on, Facebook had no way of knowing who bought.
The ads themselves were doing fine. About 3 to 5 out of every 100 impressions ended in a click, and clicks were cheap. So people were interested. What went wrong was everything after the click: on phones, three pop-ups covered the first screen, and with the tracking broken, neither platform could learn who actually bought.
We fixed the tracking before touching the ads.
On Google, we set bidding to go after purchases instead of YouTube subscriptions, and paused the old campaign that had the fake conversions in its history. On Facebook, we got a working pixel on the site. Only after the tracking was fixed on both platforms did we rebuild the campaigns, this time optimized for purchases.
We can't show you what happened next. We don't work with this store anymore, we no longer have access to its accounts, and we don't publish numbers we can't check.
A fake 22x costs real money.
With 22x in the report, you raise the budget without thinking twice. At 1.2x, every dollar on ads brings back $1.20 in sales, and out of that you still pay for the product, shipping, and fees. The ads only break even if everything else an order costs stays under a sixth of the price. This store had told us it kept about 40 cents of every sales dollar after product and other costs, so its ads needed to bring back at least $2.50 for every dollar just to break even. The one campaign with honest numbers made 1.2x.
That's the real damage of a fake number: it gets you to put more money exactly where you're losing it. A real 1.2x is bad news, but it's news you can work with.
Run these checks on your own accounts.
Put the conversions in Google Ads next to the orders in your store for the same month. Google only counts orders that came from its ads, so you should see fewer there than in your store. If you see more, the account is counting something twice, or counting something else.
Then divide conversions by clicks and multiply by 100. That's how many orders the campaign claims for every 100 clicks. Compare it with your store's own conversion rate, the orders you get per 100 visits. The campaign above claimed about 67. If yours is way above your store's normal rate, check what it's counting before you touch the budget.
In Google Ads, go to Goals, then Conversions, then Summary, and check which actions are marked Primary. For a store, the primary one is the purchase, with the order value attached. Add to cart can stay as a secondary action. YouTube subscriptions shouldn't be primary at all.
On Facebook, open Events Manager and check when your pixel last received a Purchase event. Then, in Ads Manager, check whether your sales campaigns are optimized for purchases or for add-to-carts. Finally, open Google's Tag Assistant and Meta's Test Events (in Events Manager), connect both to your site, and then place a test order yourself. The purchase should show up once in each, with the right amount.
Written by the HIGH-UP ADS team: Oana, Lorena and Ștefan.


