For two weeks Google has seen fewer orders than your shop. On 26 August a script went silent, not your revenue.

What has been different since 26 August

On 26 August 2026 Shopify switched off script tags on the thank-you page and the order status page for shops without Plus. The Shopify documentation on checkout.liquid names the date explicitly. Plus shops had the same cut on 28 August 2025, together with checkout.liquid and additional scripts.

Your shop did not notice. The checkout works, orders arrive, the confirmation email goes out. Nothing broke that a customer could see. What broke is the part nobody sees: the small script that told Google Analytics, Google Ads and Meta after the purchase that an order had just happened.

In a great many shops that is exactly where the purchase event lived. An app added it years ago, an agency set it up once, nobody has looked at it since. We announced this in August. Now the question is whether it hit you and how you can tell.

Why it looks like a drop in sales

The gap looks different in every tool. That is why it gets mistaken for something else for two weeks.

Google Ads shows a slow decline. The standard conversions column assigns every order to the day of the click, not the day of the purchase. The Google Ads help says so. Someone who clicked on 20 August and bought on the 27th would have been counted on the 20th. If the report fails on the 27th, the order is missing retroactively on the 20th. The gap smears across the days before 26 August and looks like a gently falling trend.

Meta shows a step down. Meta counts by the time of the event. There the curve breaks on 26 August. If only the browser pixel was reporting and no server report came with it, it drops to zero.

Google Analytics shows visits without a purchase. The purchase rate falls, the visitor count stays. That looks like a problem in the shop, even though the shop sells as before.

Shopify itself shows everything correctly. The orders sit in the admin with number, amount and time. That is the number you can hold on to.

Check in ten minutes whether it hit you

You do not need a developer for this. You need three series of numbers and a look at the settings.

1. Orders per day from Shopify since 1 August. In the admin under Analytics or simply as an export of the orders. Take paid orders only.
2. Orders per day from Google Ads for the same period. Use the column “Conversions by conversion time”, not the standard column. Only that column counts by the day of purchase and can be compared with Shopify.
3. Purchase events per day from the Meta Events Manager for the same period.

Put the three series side by side. Before 26 August, Google and Meta should each see a stable share of the Shopify orders, often between 60 and 90 percent. From 26 August that share must not change. If it collapses, it hit you.

Then the look at the settings: in the Shopify admin under Settings and then Customer events you see which pixels still run today. What is listed there runs. What is not listed there has been gone since 26 August. If the list is empty or only shows an app you no longer use, the case is clear.

Finally a real test purchase with a 100 percent discount code. It takes two minutes and shows you without guessing which platform still receives the order.

What to do now

There are two routes and the durable answer is usually both.

Route one: a pixel in the Customer events area. That is the replacement Shopify intends. The Google & YouTube app and the Meta app add their pixels there themselves. For Google Analytics, Tag Manager or your own events there is the custom pixel. It runs in the browser in a sealed-off area, with the same limits as before: ad blockers and Safari still cost a share of the reports. In return it is set up in an afternoon.

You see two mistakes again and again. First, the old script is copied unchanged into the new pixel without respecting the cookie banner. Inside the pixel the banner’s state applies and you have to ask for it. Second, suddenly both run: the new pixel through the app and the old script that a second app still built in somewhere. Then Google counts every order twice and your advertising looks better than it is.

Route two: report the order from the server. Shopify sends a message to an address of your choice for every order. From there a small service reports the purchase to Google and Meta, with the order number as the identifier so nothing is counted twice. This route does not depend on whether a page loaded, whether a script was allowed to run or what Shopify switches off next year. It costs more effort at the start and needs a cookie banner whose answer you pass along.

In practice it goes like this: the pixel for the events during the checkout, the server for the order itself. What that looks like for Meta with the identifier against double counting is in the Conversions API guide. What we do for Shopify shops is on the Shopify tracking page.

What cannot be recovered

The orders of the last two weeks do not arrive at Google and Meta after the fact. There is no button that reports them retroactively. Google Ads can add orders through an import, but only with the click identifier, which your shop usually did not store for the missing orders.

More important is what happens from now on. Google and Meta distribute your ad budget by the orders they see. For two weeks they saw too few and turned down campaigns that were selling. After the repair both need a few days until the numbers carry again. During that time compare against your own orders from Shopify, not against the curves in the ad account.

What you can do now

• Put the three series of numbers since 1 August side by side. It takes ten minutes and answers the most important question.
• Look under Customer events to see which pixels really run and make a test purchase.
• If the report is missing: set up pixels through the apps or the custom pixel and additionally have the order reported from the server. Order number as identifier, cookie banner answer passed along.
• If you are not sure what still sends and what sends twice: the tracking check takes every route apart, with a real test purchase. Within two working days you get a report that says what is missing and what it costs to fix.