Google’s Mueller: Why LCP Fixes Often Miss the Mark

Google Search Advocate John Mueller has highlighted a new case study that explains why many Largest Contentful Paint (LCP) improvements fail to produce meaningful results. The core problem? In layouts that vary significantly from one store to another, the browser may latch onto the wrong element as the LCP candidate. This means that subsequent optimization efforts are wasted on something that was never the actual LCP element in the first place.

Published on web.dev on June 24, the case study documents a year-long Core Web Vitals overhaul at the ecommerce platform Nuvemshop. Initially, the team suspected heavy images or slow server response times were the culprits. But deeper analysis revealed that the real issue was how the browser selected the LCP element. After refining element selection, image priority, lazy loading, and edge caching, Nuvemshop saw the share of its stores with good LCP scores jump from 57% to an impressive 96%.

How the Wrong Element Gets Selected

On Nuvemshop, merchants can arrange their homepage sections in any order they choose. This flexibility leads to carousels, banners, and product grids appearing in different positions across various themes. According to the case study, carousels were present on 85% of storefronts, and that is where the measurement problem began.

The trouble stemmed from CSS transitions applied to carousels and banners. These transitions delayed when elements became visible to the browser’s LCP detection algorithm. While shoppers saw the carousel load first, the browser sometimes identified a banner further down the page as the LCP element because its visibility wasn’t delayed by a transition. As a result, Nuvemshop’s earlier optimization efforts were directed at elements that were never truly the LCP.

The investigation uncovered three main causes. Transitions were pushing detection to the wrong frame. Lazy loading was delaying images above the fold that needed to load immediately. And critical images lacked priority signals. Additionally, measurements came from both category and product pages, not just homepages, so the solution had to work across different page types.

The Three LCP Fixes

The case study includes a fourth change, edge caching to reduce latency. But the core element-detection adjustments involved three modifications to how the top of the page renders and loads. Nuvemshop applied all three changes across its main themes and page types. These adjustments are straightforward and follow common web performance best practices.

The team removed CSS transitions from the top sections so they display instantly. This allows the browser to recognize them as LCP candidates without any delay.

They also stripped `loading=”lazy”` from the first image in those top sections. web.dev’s own guidance is blunt on this point, warning against lazy-loading the LCP image because it always adds load delay.

The final piece was adding `fetchpriority=”high”` to the likely LCP image. This tells the browser’s preload scanner to grab it sooner. Nuvemshop wrapped this in validation logic so the signal fires only where the element could genuinely be the LCP candidate. Google advises reserving this attribute for one or two key images, since if everything is high priority, nothing is.

What Else Nuvemshop Reports

Nuvemshop reports that its overall Core Web Vitals pass rate improved from 48% to 72% over the year, not just the LCP figure.

The company reviewed the same Brazilian stores active in both January 2025 and 2026. For mobile visitors arriving via Google organic search, it reports an 8.9% increase in conversion rate and an 8.4% rise in cart engagement.

Keep in mind these are self-reported figures, and the shopping numbers represent a year-over-year comparison of that cohort, not a controlled test. Nuvemshop ties the improvement direction to its speed work by referencing Deloitte research commissioned by Google, which found that shaving 0.1 seconds off load time can raise retail conversion by 8.4%.

Why This Matters

Before compressing another hero image, double-check that you are optimizing the element the browser actually treats as LCP. This is especially important in template-driven or carousel-heavy layouts, where the LCP element might not be what you expect.

Our own recent coverage lands in the same spot. In a look at Core Web Vitals across CMS platforms, HTTP Archive data showed that real-world LCP tends to break when a platform is slow to let the browser discover the main image, or when it forces that image to compete with too many other high-priority resources. It is not just about page weight. Barry Pollard has also walked through how to trace an LCP problem in PageSpeed Insights before you touch any code.

The conversion link reflects Nuvemshop’s interpretation of the data. Core Web Vitals are only a minor ranking factor, and Google has emphasized relevance as more important. So a faster store is better read as a shopping experience gain than a ranking lever.

Looking Ahead

This is just one company’s self-reported experience, so the shopping gains are best read as directional. The technique itself is not new either, since web.dev has been advocating for this kind of discovery-and-priority work for years. What the case study adds is a worked example of identifying the true LCP element across layouts that vary from one store to another. That is the hard part on any customizable platform.