Debugging a Rogue Referral Problem Google Support Couldn't Explain
I said yes to something I was terrified of, and it took nine months and four different Google support consultants before the data proved me right.
When I interviewed for my in-house role, I was asked about my experience with website tracking. I said I was familiar with Google Tag Manager and could handle basic setups.
The truth was I avoided GTM. Tracking sits at the border between marketing and development, and I'd always assumed development wasn't something I had the brain for. I told my interviewer I'd need help with anything advanced. I was told I'd get it.
Six months later, the company launched a new site on a headless CMS I'd never touched. There was no one left to hand tracking to. I got the developer's documentation, rebuilt the setup myself, tested it, and it looked fine.
A week after launch, my Google Ads campaigns started failing according to Analytics, while Google Ads itself reported everything working normally.
Conversion rate from paid dropped from 20% to 1% overnight.
Time on site collapsed.
Bounce rate spiked.
Nothing about actual user behaviour had changed that fast, and I knew it.
Refusing the easy answer
I raised it with Google Analytics support.
They pointed to a shift in user behaviour. I didn't accept that, because the numbers didn't support it.
Google Ads was still crediting itself with conversions on its own dashboard, and those same conversions were showing up in Analytics as a sudden surge in organic traffic instead. That's not a behaviour change, that's traffic getting mislabelled somewhere in the pipeline.
Getting past the first "it's probably your users" response was most of the work.
I asked to be routed to the tag implementation team specifically, not another generalist reading from a script.
On that call, I found out the new site was a single-page application, something I hadn't clocked as relevant until someone said it out loud.
Reframing the search, finding the real cause
I'd already searched for our CMS name plus "tracking issue" and found nothing useful.
Once I reframed the search to "single-page application Google Analytics issue," I found a wealth of writing on the exact problem, including a detailed breakdown from Simo Ahava describing what he called the rogue referral problem: on single-page sites, Universal Analytics tags fired on client-side navigation could lose the original campaign parameters and fall back to reading the HTTP referrer instead, which meant paid sessions got silently reattributed to organic.
I implemented Ahava's fix and traffic partly recovered, but sessions were still breaking internally in a smaller way.
Another few rounds of escalation later, a tag implementation developer suggested swapping out the custom navigation trigger our CMS docs recommended for GTM's built-in history change trigger.
That was the actual fix. Paid traffic stopped bleeding into organic for good.
A note on the technical detail: this was 2022, and the fix was specific to how Universal Analytics and that particular CMS handled client-side navigation. UA is gone now, and GA4 plus modern GTM setups largely handle single-page navigation out of the box. If you're chasing a similar symptom today, don't copy this fix, the platform has moved on. What hasn't moved on is the process: don't accept "it's probably your users" as an answer when the numbers say otherwise, and change your search terms until you find someone who's hit your exact problem.
What actually mattered
None of this would have happened if I'd taken the first explanation I was handed, or stayed in the "I'm not technical enough for this" story I walked into the interview with.
The job forced the issue.
Being underqualified for it is what taught me the most.
If there's a takeaway, it's this: the moment someone senior hands you a wrong-but-authoritative answer and it doesn't square with what you're actually seeing, that's the moment to keep digging instead of writing it off as your own inexperience.
