Lab score versus field data
The number most people quote is the lab score, produced by a simulated test on a simulated connection. Field data comes from the Chrome User Experience Report and reflects real visits.
They disagree frequently, and when they do the field data is the one that matters. A site can score well in the lab and fail in the field, particularly when real users are on slower devices and networks than the simulation assumes. In India that gap is wide.
Fix in this order
Each step sets the ceiling for the next, so order matters more than people expect.
- Hosting and server response, because no front-end work fixes a slow Time to First Byte
- Images, since the largest element is usually an image and correct sizing plus modern formats does most of the work
- Render-blocking CSS and JavaScript, deferring what is not needed for the first paint
- Plugin audit, removing scripts loading on every page for a feature used on one
- Caching and a CDN, which help delivery and do not fix any of the above
Interaction to Next Paint is the hard one
INP replaced First Input Delay and is considerably harder to pass on plugin-heavy stores. It measures whether the page responds when someone taps, and the usual cause of failure is too much JavaScript executing on the main thread.
Stores that passed comfortably on the old metric frequently fail this one, which surprises people who assumed their speed work was finished.
Layout shift is the cheapest fix
Cumulative Layout Shift usually comes from images without width and height attributes, fonts loading late, and banners injected after render. It is the least technical of the three and often the quickest to resolve.
It is also the one users feel most directly, because it is what makes someone tap the wrong thing and leave.
When optimisation has a low ceiling
A multipurpose theme loading everything to support anything caps how far remediation can go. At that point the honest advice is that a rebuild is a better spend than continued optimisation, and saying so is more useful than selling another round of work.
Key takeaways
- Optimise against field data, not the lab score
- Hosting first, then images, then scripts. Caching last
- INP is much harder to pass than the metric it replaced