In this article
What the numbers mean, and which ones count
Core Web Vitals are three measures of how a page feels to use. Largest Contentful Paint (LCP) is how long the main content takes to appear. Interaction to Next Paint (INP) is how quickly the page responds when someone taps or clicks. Cumulative Layout Shift (CLS) is how much the page jumps around while it loads.
Google assesses each at the 75th percentile of real page loads, separately for mobile and desktop. So 75% of visits have to meet the target, and a page only passes if all three do.
That word "real" matters. There are two kinds of data, and they answer different questions. Field data comes from real visitors, and it is what Search Console, Shopify's web performance reports and Google's assessment are based on. Lab data, from a tool such as Lighthouse, is one simulated load on one machine. Use lab data to find problems and field data to find out whether you have fixed them.
Lab tools cannot measure INP at all, because nobody is interacting with the page. Total Blocking Time is the nearest stand-in, as it counts how long the browser is too busy to respond.
What we found on DYNAMO
DYNAMO Perfumes is an editorial fragrance store we built on Shopify, with GSAP and Lenis handling scroll animation. We ran Lighthouse against the live home page on 3 November 2026 with simulated mobile conditions. The performance score was 68, with LCP at 2.9 seconds, CLS at 0 and Total Blocking Time at 890 ms.
The pattern is common for a design-led store. CLS is excellent, because the layout reserves its space. LCP is close but over the line. Blocking time is the real issue: that much JavaScript at load suggests people on mid-range phones will wait for the page to respond.
We publish this because it is more useful than a screenshot of a perfect score, and because it shows where the effort should go. An earlier version of this store had a different problem: as we wrote in our notes on the build, mobile CLS on the quiz flow was too high at launch, because images on the result screen loaded late. We fixed that after launch. It should have been caught in QA, and it is why we now test on a real mid-range phone before launch.
Fixing LCP on a Shopify theme
On most stores the LCP element is the first big image: a hero banner on the home page and the product photo on a product page. Almost every LCP problem on Shopify comes down to how that image is loaded.
Do not lazy-load it. Many themes apply lazy loading to every image in a section, including the first one. Set the first section's image to load eagerly with fetchpriority="high", and keep lazy loading for everything lower down.
Serve the right size. Shopify's image_url filter resizes images for you, and the image_tag filter writes the srcset, sizes, width and height attributes. Use them instead of a single huge file, and give each image a sizes value that matches its layout, so phones don't download a desktop image.
Get out of its way. If the hero sits in a slider, or fades in through JavaScript, the browser cannot show it until that script has run. DYNAMO has no sliders, which helps. Show the first image with plain HTML and CSS, and animate it afterwards if you must.
Then deal with fonts and CSS. Critical CSS should arrive early, and a web font should swap in rather than hold up the page. Shopify's font_face filter accepts font_display: 'swap', and two families and a few weights are usually enough for an editorial look.
Fixing INP: apps, animation and long tasks
INP is where Shopify stores most often fail on mobile, and the cause is usually too much JavaScript. Shopify's own performance documentation lists theme code, apps, images, videos, carousels, social media feeds and analytics as features that can affect speed.
Start with apps. Go through your installed apps and remove any you no longer use, along with any code they left in the theme, because uninstalling does not always remove it. For the rest, ask whether each one needs to load on every page, and whether a review widget really has to run before the visitor has scrolled to it.
Then look at animation. Smooth-scrolling and timeline libraries such as Lenis and GSAP can look beautiful, but they run on the main thread alongside everything else. Load them after the page is interactive, animate only transform and opacity, and turn them off for people who have asked for reduced motion. On a phone, ask whether a given effect is worth what it costs. Often a simple fade does the same job.
Finally, split up long tasks. Open the Performance panel in Chrome DevTools, throttle the CPU, and record a tap on a filter or a variant selector. Anything over 50 ms is a long task, and the panel shows which script caused it. Fix the worst first.
Fixing CLS: reserve the space
CLS has the simplest fix: tell the browser how much room something will need before it arrives. The image_tag filter adds width and height, and CSS aspect-ratio handles the rest. DYNAMO's score of 0 comes from doing this everywhere.
The usual offenders are app widgets that appear after load, such as review stars and announcement bars. Give them a container with a minimum height. Web fonts can cause a smaller shift when they swap in, so choose a fallback font with similar proportions.
Dynamic views need the same care. Our own lesson on DYNAMO came from a quiz result screen, where images loaded after the layout had settled. Any step that swaps content in, whether a quiz, a filter or a variant change, needs its image dimensions known in advance.
What not to bother with
Skip speed-booster apps that promise a score in one click. Many work by delaying scripts until the visitor interacts, which can flatter a lab score without helping a real person, and they add another script to maintain. Fix the underlying cause instead.
Do not chase a Lighthouse score of 100, either. Google assesses field data against the three thresholds, so a store passing all three at the 75th percentile has reached the goal, whatever its lab score says. Past that point, your time is better spent on the things that sell: photography, product copy and checkout.
The order we work in
We start with field data, from Shopify's web performance reports or Search Console, to see which metric fails and on which template. A product page failing LCP and a collection page failing INP are separate problems.
Next we remove things. Unused apps and leftover scripts are free wins, so they come before any optimisation. Then we fix the LCP image on each main template, reserve space for anything that shifts, and only then start on JavaScript.
We test on a real mid-range Android phone, because a fast laptop hides most of these problems. Lighthouse confirms the direction of each change. Then we wait for field data to catch up, which takes up to 28 days, since the Chrome data behind Search Console is a rolling window.
Budget for this as part of a build rather than an afterthought. Our ecommerce development service includes speed and Core Web Vitals work on every custom store, at £2,000–£7,000. If your existing store is failing, send us the URL and tell us what you have already tried.
Sources
Frequently asked
Quick answers
Can a Shopify store pass Core Web Vitals?
Yes. Shopify's hosting and image handling are a good base, and failures usually come from the theme, apps and animation rather than the platform. Check your own field data in Search Console or Shopify's web performance reports, since passing depends on what your store has installed.
Why is my Shopify store slow on mobile?
Usually because of JavaScript from apps and the theme, and a hero image that loads late or is too large. Phones have slower processors than the laptop you test on, so the same script takes much longer to run. Test on a mid-range phone with CPU throttling.
Do Shopify apps slow down my store?
Many do, because they add scripts to the storefront. Shopify lists apps among the features that can affect performance. Remove what you don't use, check that the code went with it, and ask whether the rest need to load on every page.
How long until fixes show up in Search Console?
Up to about 28 days, because Core Web Vitals field data covers a rolling 28-day window. Use Lighthouse to confirm a fix straight away, then let the field data catch up before you judge the result.



