How can server-side GTM make your data worse — even if it was set up “by the book”?

Why does “better tracking” sometimes feel worse?

You roll out server-side GTM for control and accuracy. Then BigQuery shows surprises: users spike, sessions inflate, unattributed traffic grows, CRM comparisons drift.

Everyone trusts the setup — so everyone looks elsewhere.

What actually goes wrong?

Most failures come from hybrid tracking. Some events go through the server. Others still fire directly from the browser.

That creates two identity systems:

• server-side FPID
• browser-generated client_id

One user becomes two. Then attribution starts choking.

The uncomfortable truth:

Server-side is not “add-on tracking”. It’s an architectural choice. The moment identities split, data quality collapses quietly.

What I lock down early

I audit event routing end to end. All critical events go through one pipeline with one identity logic. I also monitor sudden jumps in users or sessions — they’re usually duplicate IDs waving a red flag.

Want to get all my top Linkedin content? I regularly upload it to one Notion doc.

Go here to download it for FREE.