Introduction
Core Web Vitals are three measurements of how a page feels to a real visitor: how quickly the main content appears, how fast the page reacts to taps and clicks, and whether the layout jumps around. A page passes when it meets Google's published thresholds for all three. For a business owner, the useful question is not what the score is today but which specific page element is failing and what it costs to fix. Below is a plain-English way to read the numbers, the difference between lab and field data, and how to scope the work sensibly.
What LCP, INP and CLS measure
The three metrics each describe one part of the visitor experience, and each fails for different reasons. Reading the LCP INP CLS set as three separate problems, rather than one score, makes fixes much cheaper to target.
- Largest Contentful Paint (LCP) is the time until the biggest visible element, usually a hero image, banner or main heading block, has rendered.
- Interaction to Next Paint (INP) reflects how long the page takes to visibly respond after a visitor taps, clicks or types, judged across the visit rather than a single moment.
- Cumulative Layout Shift (CLS) scores unexpected movement, such as a button sliding down just as someone tries to press it.
How Core Web Vitals are judged: thresholds and the 75th percentile
Google's guidance on web.dev sets the good thresholds at the 75th percentile of page loads: LCP at 2.5 seconds or less, INP at 200 milliseconds or less, and CLS at 0.1 or less. The 75th percentile matters. It means three out of four visits should meet the threshold, so a fast result on your own office broadband does not count if a quarter of visitors on slower phones have a poor experience.
A page passes only when all three metrics are in the good range. One failing metric is enough to hold it back, which is why a single heavy chat widget or oversized banner can undo otherwise careful work.
Field data versus lab data
Lab data comes from a test run in a controlled setting, with a fixed device and network profile. It is repeatable, quick, and good for debugging because you can change something and test again. Field data comes from real visitors using real devices, connections and browsing habits. The two often disagree, and the source guidance notes that lab and field results differ.
Treat field data as the verdict and lab data as the diagnostic tool. If real-user reports show a failing INP, a lab test that looks fine does not clear the page; it only means the lab run did not reproduce the interaction that hurts people. Field data also needs enough traffic to be reported, so low-traffic pages may have no field numbers at all. In that case, use lab tests and judge by the page template rather than waiting for data that may never appear.
Core Web Vitals optimisation: fixes matched to the failing metric
When LCP is slow
Identify which element is the largest first. If it is an image, check that it is correctly sized for the screen, compressed, and not lazy-loaded when it sits at the top of the page. If it is text, look at web font loading and render-blocking CSS or scripts. Slow server response also delays everything, so check hosting and caching before redesigning anything. Sliders and autoplay video heroes are common offenders, and replacing them with one static image is often the cheapest improvement.
When INP is poor
Poor INP usually points to too much JavaScript running when someone interacts. Common causes are third-party tags, chat widgets, heavy filters, large menus and tracking scripts competing for the browser. Audit what each script does and who still uses it. Removing or deferring an unused tag costs less than rewriting code. Interactive features that genuinely need heavy scripts may need to be simplified or loaded only when the visitor asks for them.
When CLS is high
Layout shift is typically fixed with discipline rather than speed. Set explicit width and height on images and video embeds, reserve space for banners, cookie notices and ads, and avoid inserting content above what the visitor is already reading. Web fonts that swap with very different metrics can also nudge text; choosing a closer fallback font reduces that.
Free quote · reply within 24 hours
Want us to look at your rankings?
Send your site and the searches you want to win. You get a written plan and a fixed quote within 24 hours.
Hypothetical example: a services site with a slow hero
Consider an illustrative consultancy site, not a real client. Field reports show the home page failing on LCP and CLS but passing INP. Investigation finds a very large unoptimised hero photo, a cookie banner that pushes content down after load, and a late-loading font. The fixes are resizing and compressing the image, reserving fixed space for the banner, and adjusting font loading. The team leaves the scripts alone because INP already passes. That narrow scope is the point: fixing what fails, and nothing else, keeps the work small.
Budgeting the work
Costs depend on the site's platform, how many templates are affected and how many third-party tools are installed. ALL figures in the table below, including the hourly rate and every cost, are hypothetical assumptions for arithmetic only, not quotes, market averages or vendor fees.
| Line item | Assumed hours | Assumed rate (INR per hour) | Illustrative cost (INR) |
|---|---|---|---|
| Field data review and audit | 6 | 2,000 | 12,000 |
| Image and hero fixes | 8 | 2,000 | 16,000 |
| Script and third-party tag cleanup | 10 | 2,000 | 20,000 |
| Layout shift fixes | 5 | 2,000 | 10,000 |
| Retesting and monitoring setup | 4 | 2,000 | 8,000 |
| Total | 33 | 2,000 | 66,000 |
The arithmetic is 33 hours multiplied by the assumed rate. The illustration excludes applicable taxes, which must be confirmed separately, and also leaves out new content, hosting changes, a redesign, platform migration and ongoing care. Field data takes time to update after changes, so a retest right after the work shows lab improvement first and real-user improvement later.
Deciding what to fix first
Start with the templates that carry the most visits or enquiries, such as the home page, key service pages and product or checkout pages, rather than every URL. Fix the metric that fails, confirm in lab tests, then watch field reports. Be careful with expectations: good scores do not guarantee higher rankings. They are one part of page experience, and relevance, content quality and authority still carry much of the weight. The more dependable benefit is a faster, steadier page for visitors, which tends to matter most where people are filling in forms or completing purchases.
Also guard against regressions. Each new plugin, tracking pixel or marketing banner can reintroduce the same problem. Agreeing that new scripts are reviewed before they go live is a small process change that protects earlier work.
Conclusion
Before commissioning any work, check which of the three metrics actually fails on your most important pages, using real-user data where it exists, and write the fix list against that evidence only. That keeps the scope proportionate and avoids paying to improve metrics that already pass.
If you would like a second opinion on a specific page or template, you can contact NextGen Fusion to talk through what the data shows and what a sensible first round of fixes might involve.
Free quote · reply within 24 hours
Want a fixed price for your project?
Send your site and the searches you want to win. You get a written plan and a fixed quote within 24 hours.
Keep reading on SEO
- How to Redesign Your Website Without Losing SEO TrafficPlan a website redesign without losing SEO: inventory URLs, keep key content, map one-to-one redirects, and check indexing after launch.
- How to get your business mentioned in ChatGPT, Perplexity and Google AI answersHow to get your business mentioned in ChatGPT, Perplexity and Google AI answers: be easy to find, easy to understand, easy to quote and mentioned by others.
- Website Audit Checklist: Find Problems That Could Be Losing You LeadsA website audit checklist ranked by lead impact: test forms, speed, indexing and accessibility, and record evidence and an owner for every finding.

