Core Web Vitals is Google's program for measuring real-world page experience. There are three metrics, and you need to know each of them by name before you can have a useful conversation with your developer or agency.

A useful mental model: LCP is about arrival, INP is about responsiveness, and CLS is about trust. A page with a great LCP but bad CLS feels slippery. A page with great CLS but bad INP feels broken. You need all three to be in the green for both users and Google to treat your page as well-built.

We did not see a single new problem in 100 audits. We saw the same five issues, in slightly different combinations, on site after site. If your business is in India and your CMS is WordPress, Shopify, or a custom PHP setup, there is a high chance you are carrying at least three of these right now.

1. 01
   Oversized hero images. A 4 MB JPEG straight out of a DSLR, served at full resolution to a 360-pixel mobile screen. We measured hero images that were 6 to 12 times larger than they needed to be. Fix this first, it has the biggest single-shot impact on LCP.
2. 02
   Render-blocking third-party scripts. Chat widgets, heatmap tools, multiple analytics tags, ad pixels, and a popup builder all loading in the head. Each one competes for the main thread before your own code even runs.
3. 03
   No CDN, no caching, shared hosting. A WooCommerce store in Mumbai serving an Assam customer from a single origin in Texas, with no browser cache headers and a shared host. TTFB alone was north of 2 seconds.
4. 04
   Heavy theme and plugin bloat. WordPress sites running 40+ plugins, a multipurpose theme that ships with 18 fonts, and a page builder whose shortcodes ship a 1 MB JavaScript bundle for what is essentially static HTML.
5. 05
   No image lazy-loading and no responsive srcset. Every image on the page, including the ones below the fold, downloaded eagerly at full size on first paint.

Any one of those is fixable in a week. All five together is a three-month project. Knowing which ones you have is the first job.

Speed work is not glamorous, and it is not free. But the order of operations matters a lot. If you fix the wrong things first, you will spend money and not see numbers move. Here is the order we use, in priority order, with rough effort estimates.

1. 01
   Audit before you touch anything. Run Lighthouse, PageSpeed Insights, and a real WebPageTest from a Mumbai or Singapore node. Capture LCP, INP, CLS, TTFB, total bytes, and the top three render-blocking resources. This is your baseline. Without it, you cannot prove anything worked.
2. 02
   Compress and resize hero images first. Convert JPEGs to WebP or AVIF, generate responsive srcset sizes, and lazy-load everything below the fold. In our audits this alone moved LCP by 0.7 to 1.5 seconds on Indian mobile networks.
3. 03
   Defer and audit third-party scripts. Move chat widgets, pixels, and analytics to load after the DOM is ready, or load them on user interaction. Kill any tag that has not been reviewed in the last six months.
4. 04
   Add a CDN and turn on real caching. Edge caching of static assets is non-negotiable in 2026. Pair it with a sensible browser cache policy and a full-page cache for non-logged-in visitors.
5. 05
   Code-split and defer non-critical JavaScript. If your framework ships 800 KB of JS for a marketing page, you are paying for an app you do not have. Trim, split, and defer.
6. 06
   Subset fonts and set font-display to swap. Two weights of one font, woff2, preloaded, is plenty for a brochure site. Stop shipping five weights of three families.
7. 07
   Add preconnect and preload hints for your LCP element. This is the polish step, not the foundation. It only matters once steps 1 to 6 are in place.

If you only have time for one weekend of work, do steps 1 and 2. They will get you more LCP improvement than the other five combined.

A few things we saw teams do that actively made things worse, in roughly the order teams fall into them:

• Chasing a perfect Lighthouse score instead of fixing real-user data. Crank the audit score to 100 with a dev build, ship it, and watch CrUX get worse. Lighthouse lab data is a starting point, not the destination. CrUX and a Real User Monitoring tool are where the truth lives.
• Throwing a faster server at a frontend problem. Moving from a host to a host will not save you if your hero image is still 4 MB and your chat widget is still render-blocking.
• Measuring only on Wi-Fi. Indian users are overwhelmingly on 4G, often on mid-range Android devices. Test on a throttled 4G profile with a mid-tier CPU slowdown. If it is fast there, it is fast.
• Treating speed as a one-time project. Sites regress. A new plugin, a new ad pixel, a holiday banner, any of these can quietly push you back over the threshold. CrUX is a moving target and you need a quarterly check-in.
• Optimizing the homepage first. The homepage is rarely the money page. Your top 5 paid landing pages and your checkout flow are almost always a better place to start.

Slow pages are not a developer aesthetic preference. They are a tax on every rupee of ad spend, a brake on every SEO project, and a quiet leak in every conversion funnel you have. The good news is that Core Web Vitals failures are unusually fixable compared to most technical debt. The five offenders we saw are well-understood, well-documented, and battle-tested. There is no exotic research required. There is only the discipline to audit, prioritize, and ship the fixes in the right order. The businesses that win the next two years in Indian digital will not be the ones with the prettiest design. They will be the ones whose pages feel instant on a ???12,000 phone on a 4G connection in tier 2 and tier 3 cities. That is the bar, and it is lower than you think. Most of your competitors are not even close. Immediate Action Step: Open PageSpeed Insights and WebPageTest right now, run them against your top 5 paid landing pages from a Mumbai node, and log the LCP, INP, CLS, and TTFB for each. Pick the single worst-performing page, and ship two fixes this week: convert its hero image to WebP with a responsive srcset, and defer every non-critical third-party script. That is one focused afternoon of work, and it is almost certainly worth more than your last ad creative.