Fast page speed does not mean fast experience
Lighthouse gives you a 95. Your users still feel the site is slow. These are measuring different things.
A store scores 95 on Lighthouse. The team considers performance solved. Conversion rate doesn't improve. Session recordings show users pausing, hesitating, re-clicking.
The score measures page delivery. The user experiences page interaction. These are different systems.
Page delivery is how quickly the browser renders pixels. Lighthouse is excellent at measuring this. Modern CDNs and frameworks have largely solved it for most stores.
Page interaction is how quickly the site responds to user intent. Click a variant — does it update instantly or lag 400ms? Add to cart — does the button respond immediately or wait for an API round-trip? Open a filter — does the page reflow or stay stable?
Lighthouse doesn't measure any of this.
The third layer is cognitive speed — how quickly the page communicates what the user needs to know. Inconsistent pricing across the page, ambiguous stock signals, and conflicting delivery information all slow decision-making in ways no performance tool detects.
In practice, the stores with the best conversion rates at a given traffic level aren't the fastest by Lighthouse metrics. They're the ones where interaction latency is under 100ms and cognitive load is minimised. A 70-score page with instant interactions and clear information architecture outperforms a 98-score page that lags on variant changes and buries availability in a tab.
Performance is a system property, not a score. Measuring the score instead of the system explains why most performance work has no commercial impact.