The client swore the tracking was fine. It wasn't. It never is.
I have never once opened an account where the conversion tracking was exactly what the owner believed it was. Not a single time. The setup is always described to me with total confidence (“yeah, we track leads, it’s all in there”), and then I open the account and find something quietly off. Sometimes small. Often not.
So when someone hands me an account and tells me the tracking is fine, I treat that the way I treat a Google recommendation. As a hypothesis, not a fact.
Here is the uncomfortable part. Conversion tracking doesn’t break loudly. There’s no error message, no red banner, no email. It breaks silently, the numbers keep flowing, the dashboards keep looking healthy, and the bid strategy keeps optimising toward something that may not be what you think it is. That is the danger. A broken setup that reports zero would get caught in a week. A broken setup that reports plausible-looking numbers can run for months.
A bid recommendation made against a broken conversion setup is worse than no recommendation at all.
The ways it goes wrong are not exotic. They are the same handful, over and over. A supposedly lead-based account turns out to have a value-per-lead test still feeding the bid strategy. Primary and secondary conversion actions get reshuffled, so the algorithm is now bidding toward an action nobody meant to prioritise. Enhanced conversions quietly degrade because of a consent change or a hashing failure. Value rules misfire. Several overlapping conversions inflate the count so every report reads better than reality. A GA4-imported action gets re-enabled by accident. A lead-form action runs on a 1-day-click attribution window and undercounts everything with any real consideration time.
That’s the list. None of it announces itself. All of it changes what your account is actually optimising toward.
So before I touch a single bid or budget, I go to Tools and Conversions and I read it like I don’t trust it, because I don’t. Which actions are primary, which are secondary. What value source each one pulls from. The last-seven-day status on each. And the question that catches the most: is any campaign running value-based bidding when the owner just told me this is a pure lead-gen account?
That last one is where I stop being polite about it. If the client tells me they’re lead-gen and I can see value-based bidding switched on, that isn’t a detail to note in the appendix. It’s the first finding, and I put the contradiction straight on the table. “You told me this is lead-gen, but the account is bidding on value. One of those is wrong. Which is it?”
I used to soften these. I’d write myself a little story to smooth it over, “this might be because they ran a test once,” and move on. I’ve stopped doing that. The story is almost always wrong, and worse, it lets the contradiction survive. Now I just ask the question, get the real answer, and audit from there. (You’d be amazed how often “oh, yeah, someone set that up last year and forgot” is the whole explanation.)
This is the least glamorous work I do. Nobody hires an agency to admire a clean Conversions tab. But every clever bid strategy, every budget reallocation, every audit finding I’ll ever write, all of it sits on top of whether the account is counting the right thing. Get that wrong and everything downstream is confidently, expensively wrong.
If you take one thing from this: don’t trust your tracking because it looks fine. Look at it because it looks fine. That’s exactly when it’s been lying to you the longest.