Your PageSpeed Score is a Lie

Client came to us frustrated. "Our PageSpeed score is 92. Google says we're fast. But customers keep complaining."

She wasn't wrong. The score was green. Core Web Vitals looked healthy. On paper, the site was flying.

So we did something old-school: we watched real people use it.

Within five minutes, we saw the problem. The homepage loaded quickly—technically. But when users clicked the navigation? A 400ms delay before it responded. When they hovered over product cards? The animation stuttered. The page was "done" according to metrics, but JavaScript was still hydrating in the background.

The site was fast to load. It was slow to use.

We made three changes:

Skeleton screens everywhere. Users saw something immediately, even if data was still loading. The perceived wait time dropped dramatically.

Interaction-first JavaScript. We prioritized making buttons clickable over loading analytics scripts. Users don't care if your tracking pixel fired 200ms earlier.

Optimistic UI updates. When you add to cart, we show the cart updating instantly—before the server confirms. In 99.9% of cases, the server agrees anyway.

Our PageSpeed score dropped to 87. User satisfaction jumped 23%.

The lesson: metrics are useful, but they measure machine time. Your customers experience human time. Sometimes making a site technically slower makes it feel faster.

I'm not saying ignore your scores. But don't optimize for robots when your customers are people.