Lighthouse says 98. Your patients say the site takes forever. Both are telling the truth.
The staff tested on office wifi with this year’s iPhones. The patients are on three-generation-old Androids with cracked screens, riding one bar of 4G in a parking garage. Lab scores measure a simulation; your revenue arrives on real devices.
Where the gap hides
Third-party tax
Chat widgets, booking embeds, tag managers, tracking pixels — each “lightweight” script negotiates its own connections on a weak network. The lab test forgives them; the parking-garage 4G does not.
The device gap
JavaScript that parses instantly on a flagship takes multiples longer on the median phone your customers actually own. If your audience skews older or budget-conscious, your real median device is slower than any default simulation.
The unmeasured moment
The homepage scores 98 — but the slowness lives in the booking flow, the insurance-check form, the gallery. Users judge the whole journey; lab tests judge one page load.
What to do instead of chasing green
- Trust field data where it exists (Chrome UX data in PageSpeed’s top panel — “what real users experienced”); where traffic is too low, add lightweight real-user monitoring.
- Audit third parties annually: list every script, ask what each earns, delete the freeloaders. The fastest code is the code you remove.
- Test the actual journey on an actual cheap phone on cellular — a €150 Android is the most honest testing tool an owner can buy.
- Fix in revenue order: the booking path before the blog, the forms before the galleries.
We hold our own site to this standard — and we still log it publicly when a metric looks better in the lab than in life. Green scores are a starting line, not a finish.
Questions people ask
Why does my site feel slow despite good PageSpeed scores?
Lab scores simulate one standardized visit. Real users bring older devices, weaker networks and journeys through booking flows the lab never tests. Third-party scripts usually explain most of the gap.
What is the difference between lab data and field data?
Lab data is a controlled simulation run on demand; field data aggregates what real visitors actually experienced. When they disagree, field data is the truth that matters.
What single step most improves real-world speed?
Usually removing third-party scripts that don't earn their cost - each widget and pixel taxes every visit on every weak connection. Audit them yearly and delete freeloaders.
Why does my site feel slow despite good PageSpeed scores?
Lab scores simulate one standardized visit. Real users bring older devices, weaker networks and journeys through booking flows the lab never tests. Third-party scripts usually explain most of the gap.
What is the difference between lab data and field data?
Lab data is a controlled simulation run on demand; field data aggregates what real visitors actually experienced. When they disagree, field data is the truth that matters.
What single step most improves real-world speed?
Usually removing third-party scripts that don't earn their cost - each widget and pixel taxes every visit on every weak connection. Audit them yearly and delete freeloaders.
Scenarios are composites built from real, common situations — not client stories. We publish no metrics we cannot prove.