Latest · Ad platforms

Your Ad Platform's Conversion Count Is a Guess

September 2, 2026

Google and Meta are reporting conversions nobody actually made, then dressing them up as clean whole numbers so you stop asking questions. That's the claim, plain and simple. If your Google Ads dashboard says 47 leads last month, a real chunk of that 47 is modeled. Inferred from patterns. Not tracked from an actual human filling out an actual form.

None of this is new, exactly. Modeling has been quietly patching these platforms for years, filling holes left by cookie loss, iOS privacy changes, and consent banners nobody reads. What's new is that automated bidding now treats those modeled numbers as its only feedback loop. You handed the wheel to a system that's navigating off a map it partly drew itself.

Here's what that means for you. Smart Bidding and Advantage+ chase whatever your pixel calls a conversion. Set your conversion action to "form submit," and the algorithm chases form submits, real or modeled, with zero interest in telling the difference. It has no idea if that form submit turned into a toured unit or a signed lease. It just sees the number go up and steps on the gas.

That's the gap you have to manage yourself. The ad platform's job is to report what happened on its own turf. Your PMS's job is to report what actually happened to a resident. Two different systems answering two different questions. Only one of them cares whether anyone actually leased.

So here's the move. Once a month, pull raw lead counts from your CRM or PMS by source, and set them next to what Google Ads and Meta claim for the same period, same campaigns. Don't expect them to match. Expect Meta to run higher, sometimes a lot higher, since its attribution window and its modeling are the most generous in the business. That's not the question that matters. The question is whether the gap stays put or keeps growing. A steady gap month over month means the modeling, whatever it's doing, is at least doing it consistently. A widening gap means something broke: maybe your tracking, maybe the platform's math. Figure out which before you touch budget based on a platform-reported cost per lead.

One thing not to do: turn off conversion modeling or strip your pixels down to only "verified" events. It feels responsible. It also starves the bidding algorithm of volume, and Smart Bidding running on scraps performs worse than Smart Bidding running on modeled data with some noise in it. You're not fixing the estimation problem. You're trading it for a worse, confirmed one.

What you can trust from the ad platform: relative trends inside the same account, using the same conversion definition over time. Cost per lead climbing from $140 to $210 over eight weeks, on a tracking setup that hasn't changed, that's real signal, even if the underlying counts have modeled events baked in. What you can't trust: absolute lead counts as a stand-in for your PMS, and any comparison across platforms, because Google and Meta don't model the same way or over the same windows. You're comparing two different guesses and calling it math.

The habit worth building is simple. Treat the ad dashboard like a speedometer, not an odometer.

It tells you if you're speeding up or slowing down. It does not tell you how many miles you've driven.

For that, go to the PMS. Every time. No exceptions, no matter how clean the platform's number looks on a Monday morning report.

Source: