Your Numbers Are Lying to You (And AI Won't Tell You)

Your conversion report can look great and be completely wrong.

That's the uncomfortable lesson from Alina Popa's masterclass at MeasureSummit last week, and I've watched it play out more times than I'd like to admit.

Here's a real example from her session.

An ad account's conversions grew from 1,067 to 4,677 over two years. More than 4x. Looks like a win. Except actual purchases fell about 23% year over year.

What happened? Over time, things like YouTube subscriptions got counted as conversions. By August 2026, they represented 72% of the conversions reported in Google Ads.

This has been a pet peeve of mine for about 20 years, yet it keeps happening.

Why? It often comes down to inexperience. You know, a junior team member who takes the numbers as facts and doesn’t question the source.

You’ve probably seen this too. A marketing agency bragging about an 88% conversion rate. That type of result doesn't happen, even on Amazon Prime Day.

You know who doesn’t trust that conversion rate? The client who says "are you nuts? Our revenue went down."

Skepticism comes from experience, not software.

Which brings me to something Alina taught that I really like. She uses simple Google Ads scripts to record what changed in an account. No AI involved.

One of those scripts caught a change Google's auto-applied recommendations made at 5:30 AM on a Sunday. Nobody on the team noticed. But ad performance would have tanked if the alert didn’t trigger a quick fix.

Could AI have caught that? Maybe days (or months) after the fact. Once performance was down and you asked AI to investigate what happened.

But not proactively.

A veteran marketer or analyst will generally find that the old-fashioned way of working with data (like from 2 years ago) gives you a straight answer faster than anything else.

And as AI becomes super-intelligent, it’s easier than ever to get help developing these scripts or “old school” hacks that give you deterministic results.

That brings me to one final insight. Alina’s rule about using AI responsibly with measurement is worth writing down: AI proposes the solution, and a human authorizes the action.

In other words, use AI in your day-to-day, but don't let AI approve its own work.

Keep a human in the loop.

To Your Measured Success!

–Jeff Sauer

CEO of MeasureU

P.S. Want the full session, including Alina's scripts and two worked example accounts? Get the recording of her Quiet the Noise masterclass here.

New videos this week from MeasureU

• Google Called It Data Studio Again. What That Means for Your Reports.

Coming Up This Week at MeasureU

Mastermind (MeasureMastermind)

Tuesday, October 13, 1:00 PM - 2:00 PM EDT

Your weekly power-up session! Live, interactive group coaching where our incredible MeasureU coaches answer your burning measurement questions, share strategies, and help you level up.

This call is for MeasureU Pro Members

Brain Gainz: Optimize GHL In 10 Minutes Using MCPs & Flows

Thursday, October 15, 11:00 AM - 12:00 PM EDT

JJ Reynolds is joining us for a hands-on Brain Gainz session on getting GoHighLevel set up so it tracks what actually matters. JJ will show you how to self-audit your GHL account for paid ads leaks, turn on conversion tracking, and build reporting around your goals, so you know exactly what to focus on each day. He'll walk through which way to access the MCP, why you shouldn't rely on the out-of-the-box conversion tracking, what your AI should be doing for you, and how to keep everything pointed at your goal.

This call is for MeasureU Pro Members