We checked tracking on 245 Shopify stores: 57 of the 195 we could observe sent duplicates

On October 25, 2026 we loaded 245 Shopify stores in a browser and recorded the tracking requests they sent. We wanted to see how often the basic events reach GA4, Meta and TikTok, and whether stores running attribution tools look any different.

The short version: 195 of the 245 stores sent browser events we could see, and 57 of those (29%) sent at least one duplicate page-level event, mostly in GA4. Missing product views were less common. Stores running attribution tools weren't clearly cleaner, but this data isn't a test of those tools.

This is one visit per store from one location, and the sample isn't random. The limits are listed below, and they matter.

Summary

Measure	Stores	Share
Loaded successfully	245
Sent visible Meta, GA4 or TikTok events	195 of 245	80%
At least one duplicate event	57 of 195	29%
Page view but no product view	25 of 166	15%

How we did it

We used a headless Chromium browser on a server in Japan, with a desktop user agent and English settings. Each store got a new browser session:

1. Load the home page and wait 10 seconds.
2. Load the first product in the store's public product feed and wait 10 seconds. This worked on 211 stores.
3. Record every request from the pages, their iframes and their workers.

We didn't click anything, add to cart, touch cookie banners or send any events. We only watched what each page sent by itself.

The sample. 96 of the stores came from Meta's Ad Library, so they were running Meta ads in October 2026. The other 149 came from a hand-made list of well-known direct-to-consumer brands. The sample leans toward larger brands and doesn't represent all Shopify stores.

What counted as a duplicate. The same platform, the same pixel or measurement ID, and the same page view or product view event, sent two or more times in one page load. Where Meta or TikTok sent an event ID, hits sharing that ID counted once. We found no event ID on the GA4 hits, so every repeat counted.

What counted as a missing product view. On a product page, a platform received the page view but no product view (ViewContent or view_item).

Duplicates by platform

Platform	Stores with events	Duplicates
GA4	183	51 (28%)
TikTok	63	8 (13%)
Meta	51	3 (6%)

GA4 accounts for most of the total, but the counting rules differ by platform, so don't read this as GA4 being worse. Nearly all the repeats were doubles: of the 128 repeated events we logged, 116 were sent twice and 12 three times.

We looked at three home pages by hand:

• One had a single GA4 tag sending page_view twice in the same session.
• One had two separate GA4 installs on the page, each with its own visitor ID.
• One had two Meta pixel IDs, each sending PageView twice with different event IDs.

These are examples, not a breakdown of causes.

Sending to more than one measurement ID was common. 45 of the 183 stores with GA4 (25%) did, which can be deliberate. Those stores were more likely to have a duplicate on one of their IDs: 18 of 45 (40%), against 33 of 138 (24%) with a single ID. That may mean several tags were installed separately, but most GA4 stores with a duplicate (33 of 51) had one ID.

Why it matters. If both hits reach the property, GA4 counts two page_view events, so reports built on event counts run high. Meta and TikTok use event IDs to remove repeats, so hits with different or no IDs may each count. We didn't look inside any store's reports, so we can't say how much this changed anyone's numbers.

Missing product views

Platform	Stores checked	No product view
TikTok	46	15 (33%)
Meta	25	6 (24%)
GA4	160	10 (6%)

Stores checked means we loaded a product page and the platform received a page view on it. Overall, 25 of 166 stores had at least one platform with this gap.

TikTok and Meta had the highest rates, in small groups. Some of these stores may send product views from a server, which we can't see.

How stores with attribution tools compared

We detected tools from network requests only. Stores can appear in more than one row.

Requests seen to	Duplicates	No product view
Triple Whale	15 of 57 (26%)	8 of 50 (16%)
Elevar	15 of 53 (28%)	5 of 43 (12%)
Northbeam	10 of 28 (36%)	2 of 20 (10%)
Any of these, or Hyros	29 of 109 (27%)	12 of 90 (13%)
None of these four	28 of 86 (33%)	13 of 76 (17%)

Our reading: these numbers don't show a real difference. A gap of a few points in groups this small is noise, and only one Hyros store turned up, so it says nothing about Hyros.

This isn't a test of the tools. Stores using them may differ in size and setup. A tool that runs only on a server wouldn't show up, so its store would land in the last row. We also didn't measure what these tools are bought for, which is attribution and server-side delivery.

Limits of this data

• One visit per store. Tags can behave differently between visits. One store that showed a duplicate in the scan didn't show it when we checked it again.
• A headless browser in a data center in Japan. Stores may treat that traffic differently through bot detection, or show different consent rules by region.
• Browser requests only, for about 10 seconds a page. Server-side events and anything firing later don't show up.
• Requests sent, not events received. We recorded what each page sent, not whether the platform accepted it.
• No purchases. We didn't add to cart or check out, so this says nothing about Purchase events.
• Not a random sample. The list leans toward larger brands and stores running Meta ads.

50 stores showed no events, and Meta appeared on only 51 stores even though 96 came from the Ad Library. We can't tell how much of that is server-side sending, consent, bot detection or timing. So neither number tells you how many stores use a platform.

We aren't publishing store names, since this is a snapshot of one visit and not a judgment of any store.

How to check your own store

You don't need a tool for a first look. Open a product page in Chrome, open developer tools, go to the Network tab and reload the page.

• GA4: filter for collect. Each request shows a tid (your measurement ID) and event names in en. Some requests batch several events in the request body, so check that too. Count page_view and view_item for each ID.
• Meta: filter for facebook.com/tr. Each hit has an id, an ev (the event name) and usually an eid (the event ID). Count PageView hits for each ID, treating hits with the same eid as one.
• TikTok: filter for analytics.tiktok.com and look for the same events.

If you find a repeat, check whether the same tag is installed twice. Look in the theme code, Google Tag Manager and any tracking apps.

Our guide to attribution software for Shopify covers choosing a tool, and the server-side tracking guide covers the events this study can't see.

Using this data

You're welcome to cite these numbers. Please link to this page or cite the DOI below, and mention the sample and the single visit, since the figures don't describe all Shopify stores. If you have questions about the method, use our contact page.

The totals and the script that built them are on GitHub and archived on Zenodo (DOI: 10.5281/zenodo.23993913). The data is under CC BY 4.0 and the script under the MIT License. Store names aren't included.