Shopify Conversion Tracking: Every Method and Its Failure Mode

The rest of this post takes each method in turn, then deals with the two problems every store hits once tracking works: double counting, and a conversion value that does not match the money.

• Native channel apps

  Google & YouTube, Facebook & Instagram and TikTok install their own tracking and send purchases to their own platform. Least work, least control, and each app reports only to its own owner.

• App pixels

  A third party app registers its own pixel and receives customer events. Fine for the job the app does, and invisible to you: you cannot read what it sends without asking the vendor.

• Custom pixels

  Your own code subscribing to the Customer Events API in Settings, Customer events. Full control over which events go where, inside a sandbox with no access to the storefront DOM.

• The order itself

  The Admin API or an order webhook hands you the paid order with its total, its currency and its status. No browser involved, so nothing can block it, and refunds arrive too.

Most stores end up with one of the first, one or two of the second, and neither of the last two. That combination is exactly the one where every platform reports its own number and nobody can reconcile them against the order list.

The Customer Events API is an event bus. Standard events fire on every page of the store including checkout, and your custom pixel subscribes to the ones it cares about. This is the replacement for everything that used to live in a script box, and it is the method worth understanding even if you never write one yourself.

analytics.subscribe("checkout_completed", (event) => {
  const checkout = event.data.checkout;

  // Send what you need, where you need it. This runs in a sandbox:
  // no access to the storefront DOM, no jQuery, no theme variables.
  fetch("https://track.yourdomain.com/e", {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({
      event: "purchase",
      order_id: checkout.order.id,
      value: checkout.totalPrice.amount,
      currency: checkout.currencyCode,
    }),
  });
});

• It runs in a sandboxA custom pixel cannot read the storefront DOM or the theme's variables. Anything it needs has to come from the event payload, which is why a tag that used to scrape the page has to be rewritten rather than pasted.
• It covers checkoutThis is the reason it exists. Theme scripts never ran on Shopify's checkout, so checkout tracking before this API was either a Plus-only checkout.liquid edit or nothing at all.
• It is still the browserThe subscription fires in the shopper's browser under their consent decisions and their blockers. It is a better place to put a tag, not a way around the reasons tags go missing.

Once tracking works on four platforms, their conversion counts add up to more purchases than your order list has. Two different problems produce that, and they need different fixes.

• The same order, claimed by several platformsEvery platform counts a purchase it can claim under its own window, and a shopper who touched three of your campaigns satisfies three of them. This is not fixable inside the ad accounts. It is fixable by counting the order once, from the order list, and then choosing a model.
• The same order, sent twice to one platformA browser pixel and a server-side event for the same purchase are two conversions unless they carry the same event ID. If you run the Meta app and your own Conversions API feed, or a channel app and a custom pixel, check the deduplication before you trust either number.

The order of operations matters here. Fix the duplicates inside each platform first, because that is a configuration problem with a right answer. Then stop adding platform conversion counts together at all, because that one is arithmetic and no configuration will save it.

A conversion without a value teaches the platform that all purchases are equal, which is why value based bidding on a flat number performs like no bidding strategy at all. Three details decide whether the value you send is worth anything.

• Send the order total with its currency. A multi-currency store that sends amounts without the currency code is teaching four platforms that a 4,000 unit order and a 40 unit order are the same size.
• Decide on gross or net and keep it. Shipping and tax in or out is a choice; changing it mid quarter makes your own ROAS history unreadable.
• Net the refunds. A browser pixel fires once, at the moment of the order, and knows nothing about the return three weeks later. Only a method that reads the order can send that correction, which is why refund-heavy categories need the order as the source of truth rather than a pixel.

1. Inventory what is running. Settings, Customer events lists every app pixel and custom pixel. Open the channel apps too. Most stores find one tracking method they had forgotten.
2. Pick one owner per platform. Either the channel app tracks Meta or your own pixel does, not both, unless they share an event ID.
3. Move checkout tracking into a custom pixel, since a theme script has never run on checkout and the script boxes are being retired.
4. Put UTMs on every link you control, because no conversion method invents a source that was never in the URL.
5. Send the order value with its currency, and decide gross or net once.
6. Read the order server-side as the number you reconcile everything against, and let the paid order be what goes back to the ad platforms with its value and its refunds.

Steps one to five are housekeeping and you can do them yourself this week. Step six is the structural one: it is the difference between a store that reports what its pixels caught and a store that reports what it sold.

Next to this one: what Shopify's own attribution measures and Shopify server-side tracking. For other shop systems, WooCommerce conversion tracking and Shopware server-side tracking. On the product side, the Shopify integration, conversion tracking software and the Meta Conversions API. If your counts disagree, start with why GA4, Meta and Google never match.