Website Speed Fear-Mongering in the Blink of an Eye

A website vendor told one of our dealers last month that our tag was slowing his website down, and that theirs was better because it’s built in. “It’s integrated.”

Rather than argue, I had Google measure it.

Eighteen dealership websites, all built by the same website vendor, so the platform is held constant. Nine of them running our PureInfluencer tag, nine without it. Mobile only. All pulled the same afternoon using Google’s own PageSpeed Insights.

Every measure came back statistically indistinguishable between the two groups. Load time, responsiveness, layout stability, blocking time, overall score. Nothing separated them.

The one that matters most in this argument is layout stability. Cumulative Layout Shift is Google’s measure of how much a page jumps around while it loads, and it is precisely the thing an overlay would be blamed for. Sites with our tag came in at 0.043. Sites without it came in at 0.044. Google’s threshold for good is 0.1, so both groups clear it with room to spare, and the gap between them is one thousandth of a point.

For scale on the load time numbers: the difference between the two group averages was 0.22 seconds. A single human blink takes between 0.1 and 0.4 seconds. The entire gap is one blink.

Here’s the part I didn’t go looking for.

The free PageSpeed report everyone screenshots gives you two different numbers. One is a single simulated test run on a pretend phone in a data center. The other is real measurements from actual Chrome users on their actual devices and networks over the last 28 days.

Across these sites, those two numbers had almost no relationship to each other. Rank them by the simulated test and rank them by what real shoppers experienced, and the two lists agree at a correlation of 0.02. That is functionally zero.

The site with the worst simulated result in the entire group was the fastest site in the group for real people. Under a second. The only genuinely slow site by real-world measurement scored right in the middle of the pack on the simulated test.

So when someone shows a dealer a red score and tells him his website is broken, it’s worth asking which of the two numbers he’s looking at, because one of them is measuring his shoppers and one of them isn’t.

I’m not arguing speed doesn’t matter. It does. I’m arguing that the number being used to sell against it is the wrong number and it’s being egregiously exageratted, and that most people waving it around have never looked at the other one.

Measure it yourself. Ours included. Field data, not the score. If a vendor tells you their tool is fast because it’s integrated and won’t show you the number, that isn’t an engineering answer.

And here’s what nobody’s selling against, because there’s no product to sell.

Bill Playford at DealerKnows ran an analysis on how often a shopper who asked for a price actually got one. Not a firm quote necessarily — any price confirmation at all would count.

It happened 82% of the time, over an average of nine days.

Nine days.

And these were lead handlers who knew they were being measured on it.

Website vendors and digital marketing gurus are selling dealers panic over a half-second while a nine-day broken promise sits on every vehicle page in America and nobody says a word about it.

One of those costs you a car. It isn’t the half-second.