The Ecomm Analyst

It broke a lot of tracking, and if yours survived it is because someone updated it. Checkout extensibility itself does not prevent attribution. What it removed was the mechanism most older tools relied on, which was the ability to drop arbitrary JavaScript onto the checkout and order status pages. Anything still depending on that stopped working when Shopify retired checkout.liquid and the additional scripts box.

Tools built on the newer foundations are fine. Tools that were quietly coasting on a script tag are not, and the failure is silent, which is the part that catches people.

What actually changed

Shopify replaced the editable checkout with an extension model. Custom code now runs through the Web Pixels API in a sandboxed environment, and interface changes happen through checkout UI extensions. The sandbox is the important bit for attribution. Your pixel no longer has access to the page or the window object, so it cannot read the DOM, cannot set cookies on your domain the old way, and cannot see whatever a previous script left lying around.

Instead it subscribes to a defined list of customer events, receives structured data about them, and sends that somewhere. Cleaner and more stable, genuinely better for site performance, and much less permissive.

How this shows up in your numbers

The classic symptom is a store whose sessions and add to carts look normal while purchases fall off a cliff, because the event that broke was the one firing at the end. Blended ROAS collapses overnight, nobody changed the ad account, and the first instinct is to blame the algorithm.

The subtler version is partial breakage. One payment path still reports and another does not, so you are undercounting by fifteen or twenty percent rather than everything. That can run for months without anyone noticing, especially if the store was already used to a gap between platform numbers and Shopify.

Either way, the check is the same and it takes ten minutes. Pull your Shopify order count for last week, pull the purchase or conversion count from each tool you rely on, and put them next to each other. If a tool is more than a few percent off Shopify, you have a tracking problem and not a marketing one.

What to do about it

Prefer tools that read orders from Shopify rather than only watching the browser. An attribution tool that ingests the order object through the API knows about every order regardless of what happened at checkout, and it uses the browser session for the touchpoint history rather than for counting revenue. That split is what makes a setup survive checkout changes. ThoughtMetric, which sponsors this blog, works this way, as do most of the serious tools in the category.

Beyond that, get your ad platform conversions onto server side event forwarding through the conversion APIs rather than depending on a browser pixel firing inside a sandbox. Not because server side is fashionable, but because it removes the checkout page from the chain entirely. I went into that trade off in whether you need server side tracking for accurate attribution.

There is a second reason to prefer order level ingestion, and it has nothing to do with checkout. Refunds and cancellations never fire a browser event at all, so any tool counting revenue from the page is counting gross forever. Reading the order object means those adjustments come through on their own.

Frequently asked questions

Do I need to be on Shopify Plus for this to affect me?

No. Editable checkout.liquid was a Plus feature, but the additional scripts field on the order status page was available much more widely, and plenty of non Plus stores had trackers living there. Both are gone, so a small store that pasted a snippet in from a blog post two years ago is exactly as exposed as a Plus store with a custom build.

Will my attribution tool tell me if it broke?

Some will flag a sudden drop in conversion events. Most will simply report a smaller number and let you draw your own conclusions. Assume nobody is watching this for you, and put the order count reconciliation on a monthly calendar reminder rather than trusting yourself to remember.

Does this affect UTM tracking?

Not directly. UTMs land on the storefront, which is unchanged. The risk is in whether the value captured on the landing page still reaches the purchase event, since that handoff used to happen through mechanisms the sandbox no longer permits. Tools that persist session data server side are unaffected.

Is one time passcode or accelerated checkout different?

Accelerated checkouts can skip pages that older scripts assumed a customer would pass through, so they have always been a weak point for browser based tracking. Order level ingestion handles them without any special work, which is another argument for not counting revenue in the browser.