Key takeaways
- The main product image should be eager and high priority, never lazy loaded
- Lazy loading belongs on gallery thumbnails below the fold
- Audit third party scripts before defending them; most can wait or go
- Reserve space for anything that arrives after the initial paint
- A laboratory score catches regressions; only field data describes customers
- Fix the image first, it is usually the largest single gain
When a product page fails its Core Web Vitals assessment, the cause is usually one of four things. They come up in roughly this order.
The main image is loaded late
The largest element on a product page is almost always the main image, so it determines Largest Contentful Paint. Two mistakes are common:
- The image is lazy loaded, which tells the browser to wait before fetching the single most important asset on the page
- The image is only discoverable after JavaScript has run, so the browser cannot start fetching it early
The main image should be eager, high priority, and present in the initial HTML with explicit width and height. Lazy loading belongs on the gallery thumbnails below the fold, not on the hero.
Too much JavaScript before interaction
Interaction to Next Paint suffers when the main thread is busy. On a typical shop the queue is a tag manager, a chat widget, a reviews widget, a recommendations script and a consent banner, all competing during the first few seconds.
Measure what each one costs before defending it. Scripts that are not needed for the first interaction can wait until the visitor has done something, and some can be removed entirely once someone checks whether anybody uses the data.
Layout moves after load
Cumulative Layout Shift on product pages usually comes from a short list of culprits:
- Images without dimensions, so space is not reserved
- A promotional banner injected at the top after render
- Web fonts swapping and changing the height of a text block
- Star ratings or stock messages that arrive after the initial paint
Each is fixed by reserving the space the element will occupy before it arrives.
The laboratory score is being mistaken for the field data
This one is not a performance problem, it is a measurement problem, and it causes a lot of wasted effort.
A laboratory test runs once, on one connection, usually on a fast machine. Field data comes from real visitors on real devices over a rolling window. They answer different questions.
Laboratory tests are for catching regressions before release. Field data is for knowing what your customers experience.
Use a laboratory test in your build pipeline to stop things getting worse. Use field data to decide whether there is a problem worth solving. Do not report a laboratory score as evidence that customers are having a fast experience, because it is not.
A sensible order of work
- Confirm there is a real problem using field data, not a single score
- Fix the main image loading, which is usually the largest single gain
- Audit third party scripts and remove or defer what is not needed
- Reserve space for anything that arrives late
- Re-measure, and keep a laboratory check in the pipeline so it does not regress
Frequently asked
Questions readers ask
Why does my product page have a poor LCP score?
Nearly always because the main image is loaded late, either because it is lazy loaded or because it is only discoverable after JavaScript runs. It should be eager, high priority and present in the initial HTML with explicit dimensions.
Is a Lighthouse score the same as Core Web Vitals?
No, and confusing them wastes a lot of effort. A laboratory test runs once on one connection. Core Web Vitals assessment uses field data from real visitors over a rolling window. Use the laboratory test to catch regressions and field data to decide whether a problem is real.
How do I fix layout shift on a product page?
Reserve the space every late-arriving element will occupy. The usual culprits are images without dimensions, promotional banners injected after render, fonts swapping and changing text height, and star ratings or stock messages that load separately.





