The discrepancy that turned out to be a consent banner

The store's owner came to his agency with a complaint that will sound familiar to anyone who runs paid traffic: the numbers in Shopify did not match the numbers his SEO agency was reporting. He suspected the ad platforms were off too, but he had made peace with that one.

"He knew he had problems on ads as well, but he'd already got used to the idea," says Cristian Bătrîncea, who runs Horcrux, the agency that builds and maintains the store.

That resignation is the part worth stopping on. Most merchants running Meta and Google Ads have quietly accepted that the platforms disagree with each other, that Shopify disagrees with both, and that nobody is going to reconcile them. The discrepancy becomes background noise. You learn to squint at the dashboards and go with your gut.

Cristian did not want to squint. He went looking for a tracking app.

Why this one

"I settled on it because it was one of the few apps that integrates several platforms into one tracking setup," he says, "and because it's Romanian."

Cristian Bătrîncea, Horcrux

The store was on four platforms at once: GA4 for analytics, Meta and Google Ads for acquisition, TikTok on top. Four platforms means four browser pixels, four different ways to lose an event to an ad blocker, an iOS privacy setting or a page that unloads before the pixel fires, and four separate sets of numbers that will never quite agree because they are each missing a different slice of the same reality.

Sending the events from the server instead removes most of that variance in one move. The order is recorded once, on our side, and forwarded to every platform from the same source. What each platform receives stops depending on what survived in the customer's browser.

The install went in on 15 June 2026. Within a week Cristian had seen enough to tell the client to pay for it, and the store has been on the paid plan ever since.

What changed

"We had good results in Meta ads. Not spectacular. Most of our traffic actually comes from Google Ads, and we weren't sure the results there were real either," Cristian says. "Now the data has evened out. I can see clearly where most of my orders come from, and as a bonus my ROAS went up."

The ROAS improvement is his report, not our measurement, and we would not present it as ours. The mechanism behind it is the ordinary one: when the ad platform receives every conversion instead of the fraction that survived the browser, its optimisation is working from a fuller picture, and it spends the budget differently. That is the case for accurate tracking in one sentence. The platform can only optimise toward the conversions you actually tell it about.

The part he emphasises is the reconciliation rather than the lift. Four platforms that used to disagree now agree, because they are all being told the same thing by the same system.

The thing they found on the way

Here is where the story stops being a tracking story.

A few weeks after the install, the store's reporting went strange. A large share of sessions in GA4 came through with no attribution at all. The client reported that his Shopify conversion rate had halved. His email agency said conversions were being double-counted in Klaviyo. All of it dated from roughly when the app went in, and the natural conclusion was that the new app had broken something.

We went through it with them point by point, and the app came out clean on each: one event per platform per order, no double-send, no storefront component that could touch conversion rate at all. But being able to say "it isn't us" is not the same as knowing what it is, and the cloud sat over the account for weeks.

Cristian is the one who found it. In August he traced it back to the store's cookie consent banner, which had reset at some point and taken most of the site's traffic tracking down with it. The timing had lined up with our install by coincidence and nothing more.

"I found what the problem was," he wrote. "The consent banner somehow got reset and most of the traffic was burned, and it happened around the same time as the WeltPixel install. A coincidence. But everything works fine now."

That is the outcome we would actually point to. The store had a real hole in its measurement, it had probably been there a while, and it surfaced because somebody was finally looking closely enough at the data to notice that two things which should have matched did not. Tracking you trust is what makes a discrepancy legible as a discrepancy instead of more background noise.