Discrepancy in channel attribution, hardcoded analytics vs tag manager

Question

I have set up tracking with tag manager on a website, which now is now active alongside with hardcoded analytics, both Universal Analytics (I am going, on the behalf of a client, switch from hardcoded analytics to tag manager and are sending the tracking via tag manager to another property at the moment for testing).

I expect some minor discrepancy which is fine, but there is a large difference in channel attribution for Display and Paid Search and I can't figure out why (I am looking at the same time period, made sure that tracking is active on all pages for both hardcoded analytics and tag manager and I have not tampered with the channel settings). See screenshots below:

Tag Manager

Hardcoded Analytics

Any ideas?

Edit: I now saw that the GTM version is missing the Advertising Dimensions. And that the GTM version did not have data collection from google signals enabled, will this cause the Advertising Dimensions to not populate? Because in the GTM version, all the traffic from source/medium = google/cpc is being attributed to channel = paid search, when the hardcoded version has both paid search and display from the source/medium = google/cpc.

Answer 1

You're likely have tracker objects conflict in your set up. That's the object that completely rules the context of your events/pageviews. Attribution is part of this context.

To have proper parallel tracking, you're supposed to define your tracker at least in one of the systems. In GTM, it is easier, here:

You set the tracker name there, but then you have to use the settings for all your tags.

You should disable it for when you get rid of the parallel tracking. It's tricky to use. You have to have a deep understanding of how the tracker works and what it means to have one tracker for all tags.

Try it out. If defining the tracker obj doesn't help, you will have to debug the data through GTM/Big Query. Thankfully, the client ids will be the same for both implementation, so it should be easy to compare hits across the properties.