The Ecomm Analyst

Mostly through deterministic matching, which means tying sessions together using something the customer supplies directly, usually an email address at checkout. Everything before that identification point gets stitched with first-party cookies and click IDs, which work within a single browser but not between them. The result is that cross-device tracking is real but partial, and understanding where it breaks is more useful than assuming it either works perfectly or not at all.

Why this matters for a Shopify store

The pattern is familiar. Someone sees an ad on Instagram on their phone during the day, does not buy, then sits down at a laptop that evening, searches your brand name, and orders. To Meta that is an impression with no conversion. To Google that is a branded search conversion. To you it is one customer and one order, and the interesting question is which of those touches actually did the work.

If your tooling cannot connect those two sessions, you will systematically underrate the channels that start journeys and overrate the ones that finish them.

Deterministic matching, which is the part that works

When a customer completes checkout, they hand over an email address. That is a stable identifier, and it lets a tool link this order to every prior session on any device where the same email was seen. Newsletter signups, account logins, and previous orders all provide those anchor points.

This is why returning customers are much easier to track across devices than first-time buyers. The identifier already exists. For a genuinely new customer, the email arrives at the end of the journey, so anything they did on a different device beforehand has no key to match on. That asymmetry is worth remembering when you read new-customer numbers.

Click IDs and first-party cookies

When someone clicks an ad, the platform appends an identifier to the landing URL. Meta uses fbclid, Google uses gclid, and most others have an equivalent. A tracking tool captures that value and stores it in a first-party cookie, so if the person comes back later on the same browser and buys, the original click can be credited.

The limits here are the ones everybody has been dealing with since 2021. Safari caps script-set first-party cookies at seven days. Ad blockers remove the capture entirely. Private browsing sessions leave nothing behind. And a cookie set on a phone means nothing on a laptop, because cookies do not travel between devices.

What server-side tracking changes

Sending events from your server rather than the browser makes the capture more durable, because it is not sitting in front of an ad blocker and is not subject to browser cookie lifetime rules in the same way. That improves how much data survives on a given device.

It does not solve the device-to-device problem on its own. If there is no shared identifier between the phone session and the laptop session, server-side collection just gives you two well-recorded sessions you still cannot connect.

How much stays invisible

A meaningful share of it. Anonymous browsing on a phone that leads to a purchase on a shared household computer under a different email is not recoverable by any tool honestly describing what it does. Neither is a friend recommending your product in a group chat.

This is one of the main reasons I treat attribution output as directional rather than exact, and why post-purchase surveys are worth running alongside it. I went through the wider accuracy question in how accurate e-commerce attribution tools actually are.

ThoughtMetric, which sponsors this blog, works from Shopify order data with the customer identity as the anchor, which is the deterministic approach described above. Any tool doing this well is using the same underlying mechanics, and if a vendor claims complete cross-device visibility, that is a claim worth pressing on.

Common questions

Is fingerprinting used for cross-device matching?

Some tools use probabilistic signals such as IP address, screen characteristics, and behavior patterns to guess that two sessions are the same person. It is less reliable than email matching, and it sits in contested territory under privacy regulation, so it is worth asking any vendor directly what they do here.

Does logging in help?

Considerably. Every logged-in session provides a deterministic identifier, which is why brands with strong account adoption or loyalty programs get better cross-device data than those with pure guest checkout.

Do Shop Pay and accelerated checkouts change this?

They provide the email earlier and more reliably, which helps the matching side. They can also shorten the on-site session so much that fewer touchpoints get recorded in the first place, so the net effect varies by store.

Can I measure how much I am missing?

Roughly. Compare the share of orders your tool assigns to a paid source against your post-purchase survey responses, and watch how much revenue lands in direct or unattributed. A large unattributed bucket is a reasonable proxy for journeys your tracking could not follow.