How To Repair Cumulative Structure Shift (CLS) Points

No Comments

Cumulative Structure Shift (CLS) makes an attempt to measure these jarring actions of the web page as new content material — be it photographs, ads, or no matter — comes into play later than the remainder of the web page. It calculates a rating based mostly on how a lot of the web page is unexpectedly transferring about, and the way usually. These shifts of content material are very annoying, making you lose your home in an article you’ve began studying or, worse nonetheless, making you click on on the unsuitable button!

On this article, I’m going to debate some front-end patterns to scale back CLS. I’m not going to speak an excessive amount of about measuring CLS as I’ve coated that already in a earlier article. Nor will I discuss an excessive amount of concerning the mechanics of how CLS is calculated: Google has some good documentation on that, and Jess Peck’s The Nearly-Full Information to Cumulative Structure Shift is an superior deep dive into that too. Nevertheless, I’ll give a bit of background wanted to grasp a number of the strategies.

Why CLS Is Totally different

CLS is, for my part, probably the most attention-grabbing of the Core Net Vitals, partly as a result of it’s one thing we’ve by no means actually measured or optimized for earlier than. So, it usually requires new strategies and methods of pondering to try to optimize it. It’s a really totally different beast to the opposite two Core Net Vitals.

Trying briefly on the different two Core Net Vitals, Largest Contentful Paint (LCP) does precisely as its title suggests and is extra of a twist on earlier loading metrics that measures how shortly the web page masses. Sure, we’ve modified how we outlined the person expertise of the web page load to take a look at the loading velocity of the most related content material, however it’s mainly reusing the outdated strategies of guaranteeing that the content material masses as shortly as potential. The best way to optimize your LCP needs to be a comparatively well-understood drawback for many net pages.

First Enter Delay (FID) measures any delays in interactions and appears to not be an issue for many websites. Optimizing that’s often a matter of cleansing up (or lowering!) your JavaScript and is often site-specific. That’s to not say fixing points with these two metrics are straightforward, however they’re fairly well-understood issues.

One motive that CLS is totally different is that it’s measured by the lifetime of the web page — that’s the “cumulative” a part of the title! The opposite two Core Net Vitals cease after the primary part is discovered on the web page after load (for LCP), or for the primary interplay (for FID). Which means our conventional lab-based instruments, like Lighthouse, usually don’t absolutely mirror the CLS as they calculate solely the preliminary load CLS. In actual life, a person will scroll down the web page and should get extra content material dropping in inflicting extra shifts.

CLS can be a little bit of a synthetic quantity that’s calculated based mostly on how a lot of the web page is transferring about and the way usually. Whereas LCP and FID are measured in milliseconds, CLS is a unitless quantity output by a complicated calculation. We would like the web page to be 0.1 or underneath to cross this Core Net Very important. Something above 0.25 is seen as “poor”.

Shifts brought on by person interplay are not counted. That is outlined as inside 500ms of a selected set of person interactions although pointer occasions and scroll are excluded. It’s presumed {that a} person clicking on a button may count on content material to look, for instance by increasing a collapsed part.

CLS is about measuring sudden shifts. Scrolling mustn’t trigger content material to maneuver round if a web page is constructed optimally, and equally hovering over a product picture to get a zoomed-in model for instance must also not trigger the opposite content material to leap about. However there are after all exceptions and people websites want to think about react to this.

CLS can be regularly evolving with tweaks and bug fixes. It has simply had an even bigger change introduced that ought to give some respite to long-lived pages, like Single Web page Apps (SPA) and infinite scrolling pages, which many felt have been unfairly penalized in CLS. Moderately than accumulating shifts over the entire web page time to calculate the CLS rating like has been completed up till now, the rating shall be calculated based mostly on the most important set of shifts inside a selected timeboxed window.

This implies tha if in case you have three chunks of CLS of 0.05, 0.06, and 0.04 then beforehand this might have been recorded as 0.15 (i.e. over the “good” restrict of 0.1), whereas now shall be scored as 0.06. It’s nonetheless cumulative within the sense that the rating could also be made up of separate shifts inside that time-frame (i.e. if that 0.06 CLS rating was brought on by three separate shifts of 0.02), however it’s simply not cumulative over the entire lifetime of the web page anymore.

Saying that, if you happen to remedy the causes of that 0.06 shift, then your CLS will then be reported because the subsequent largest one (0.05) so it nonetheless is taking a look at all of the shifts over the lifetime of the web page — it’s simply selecting to report solely the most important one because the CLS rating.

With that temporary introduction to a number of the methodology about CLS, let’s transfer on to a number of the options! All of those strategies mainly contain setting apart the correct quantity of area earlier than extra content material is loaded — whether or not that’s media or JavaScript-injected content material, however there’s a couple of totally different choices accessible to net builders to do that.

Set Width And Heights On Photographs And iFrames

I’ve written about this earlier than, however one of many best issues you are able to do to scale back CLS is to make sure you have width and top attributes set in your photographs. With out them, a picture will trigger the following content material to shift to make method for it after it downloads:

That is merely a matter of fixing your picture markup from:

<img src=”hero_image.jpg” alt=”…”>


<img src=”hero_image.jpg” alt=”…”
width=”400″ top=”400″>

Yow will discover the scale of the picture by opening DevTools and hovering over (or tapping by) the ingredient.

I counsel utilizing the Intrinsic Measurement (which is the precise measurement of the picture supply) and the browser will then scale these right down to the rendered measurement if you use CSS to alter these.

Fast Tip: If, like me, you’ll be able to’t keep in mind whether or not it’s width and top or top and width, consider it as X and Y coordinates so, like X, width is at all times given first.

If in case you have responsive photographs and use CSS to alter the picture dimensions (e.g. to constrain it to a max-width of 100% of the display screen measurement), then these attributes can be utilized to calculate the peak — offering you keep in mind to override this to auto in your CSS:

img {
max-width: 100%;
top: auto;

All fashionable browsers help this now, although didn’t till not too long ago as coated in my article. This additionally works for <image> components and srcset photographs (set the width and top on the fallback img ingredient), although not but for photographs of various aspect-ratios — it’s being labored on, and till then you must nonetheless set width and top as any values shall be higher than the 0 by 0 defaults!

This additionally works on native lazy-loaded photographs (although Safari doesn’t help native lazy loading by default but).

The New aspect-ratio CSS Property

The width and top approach above, to calculate the peak for responsive photographs, might be generalized to different components utilizing the brand new CSS aspect-ratio property, which is now supported by Chromium-based browsers and Firefox, however can be in Safari Expertise Preview so hopefully meaning will probably be coming to the secure model quickly.

So you could possibly apply it to an embedded video for instance in 16:9 ratio:

video {
max-width: 100%;
top: auto;
aspect-ratio: 16 / 9;

<video controls width=”1600″ top=”900″ poster=”…”>
<supply src=”/media/video.webm”
<supply src=”/media/video.mp4″
Sorry, your browser does not help embedded movies.

Curiously, with out defining the aspect-ratio property, browsers will ignore the peak for responsive video components and use a default aspect-ratio of two:1, so the above is required to keep away from a format shift right here.

Sooner or later, it ought to even be potential to set the aspect-ratio dynamically based mostly on the ingredient attributes through the use of aspect-ratio: attr(width) / attr(top); however sadly this isn’t supported but.

Or you’ll be able to even use aspect-ratio on a <div> ingredient for some type of {custom} management you might be creating to make it responsive:

#my-square-custom-control {
max-width: 100%;
top: auto;
width: 500px;
aspect-ratio: 1;

<div id=”my-square-custom-control”></div>

For these browsers that don’t help aspect-ratio you should use the older padding-bottom hack however, with the simplicity of the newer aspect-ratio and huge help (particularly as soon as this strikes from Safari Technical Preview to common Safari), it’s exhausting to justify that older methodology.

Chrome is the one browser that feeds again CLS to Google and it helps aspect-ratio that means that may remedy your CLS points by way of Core Net Vitals. I don’t like prioritizing the metrics over the customers, however the truth that the opposite Chromium and Firefox browsers have this and Safari will hopefully quickly, and that it is a progressive enhancement signifies that I’d say we’re on the level the place we are able to go away the padding-bottom hack behind and write cleaner code.

Make Liberal Use Of min-height

For these components that don’t want a responsive measurement however a hard and fast top as an alternative, think about using min-height. This may very well be for a fastened top header, for instance and we are able to have totally different headings for the totally different break-points utilizing media queries as standard:

header {
min-height: 50px;
@media (min-width: 600px) {
header {
min-height: 200px;



In fact the identical applies to min-width for horizontally positioned components, however it’s usually the peak that causes the CLS points.

A extra superior approach for injected content material and superior CSS selectors is to focus on when anticipated content material has not been inserted but. For instance, if you happen to had the next content material:

<div class=”container”>
<div class=”main-content”>…</div>

And an additional div is inserted through JavaScript:

<div class=”container”>
<div class=”additional-content”>…/div>
<div class=”main-content”>…</div>

Then you could possibly use the next snippet to go away the area for added content material when the main-content div is rendered initially.

.main-content:first-child {
margin-top: 20px;

This code will really create a shift to the main-content ingredient because the margin counts as a part of that ingredient so it’s going to seem to shift when that’s eliminated (regardless that it doesn’t really transfer on display screen). Nevertheless, at the very least the content material beneath it is not going to be shifted so ought to scale back CLS.

Alternatively, you should use the ::earlier than pseudo-element so as to add the area to keep away from the shift on the main-content ingredient as effectively:

.main-content:first-child::earlier than {
content material: ”;
min-height: 20px;
show: block;

However in all honesty, the higher answer is to have the div within the HTML and make use of min-height on that.

Verify Fallback Parts

I like to make use of progressive enhancement to offer a fundamental web site, even with out JavaScript the place potential. Sadly, this caught me out not too long ago on one website I keep when the fallback non-JavaScript model was totally different than when the JavaScript kicked in.

The problem was because of the “Desk of Contents” menu button within the header. Earlier than the JavaScript kicks in it is a easy hyperlink, styled to seem like the button that takes you to the Desk of Contents web page. As soon as JavaScript kicks in, it turns into a dynamic menu to permit you to navigate on to no matter web page you need to go to from that web page.

I used semantic components and so used an anchor ingredient (<a href=”#table-of-contents”>) for the fallback hyperlink however changed that with a <button> for the JavaScript-driven dynamic menu. These have been styled to look the identical, however the fallback hyperlink was a few pixels smaller than the button!

This was so small, and the JavaScript often kicked in so shortly, that I had not observed it was off. Nevertheless, Chrome observed it when calculating the CLS and, as this was within the header, it shifted all the web page down a few pixels. So this had fairly an affect on the CLS rating — sufficient to knock all our pages into the “Wants Enchancment” class.

This was an error on my half, and the repair was merely to convey the 2 components into sync (it might even have been remediated by setting a min-height on the header as mentioned above), however it confused me for a bit. I’m certain I’m not the one one to have made this error so pay attention to how the web page renders with out JavaScript. Don’t assume your customers disable JavaScript? All of your customers are non-JS whereas they’re downloading your JS.

Net Fonts Trigger Structure Shifts

Net fonts are one other frequent explanation for CLS because of the browser initially calculating the area wanted based mostly on the fallback font, after which recalculating it when the online font is downloaded. Normally, the CLS is small, offering a equally sized fallback font is used, so usually they don’t trigger sufficient of an issue to fail Core Net Vitals, however they are often jarring for customers nonetheless.

Sadly even preloading the webfonts gained’t assist right here as, whereas that reduces the time the fallback fonts are used for (so is nice for loading efficiency — LCP), it nonetheless takes time to fetch them, and so the fallbacks will nonetheless be utilized by the browser normally so doesn’t keep away from CLS. Saying that, if you recognize an online font is required on the subsequent web page (say you’re on a login web page and know the subsequent web page makes use of a particular font) then you’ll be able to prefetch them.

To keep away from font-induced format shifts altogether, we might after all not use net fonts in any respect — together with utilizing system fonts as an alternative, or utilizing font-display: elective to not use them if not downloaded in time for the preliminary render. However neither of these are very passable, to be sincere.

An alternative choice is to make sure the sections are appropriately sized (e.g. with min-height) so whereas the textual content in them might shift a bit, the content material under it gained’t be pushed down even when this occurs. For instance, setting a min-height on the <h1> ingredient might forestall the entire article from shifting down if barely taller fonts load in — offering the totally different fonts don’t trigger a unique variety of traces. It will scale back the affect of the shifts, nevertheless, for a lot of use-cases (e.g. generic paragraphs) will probably be tough to generalize a minimal top.

What I’m most enthusiastic about to unravel this situation, are the new CSS Font Descriptors which let you extra simply modify fallback fonts in CSS:

@font-face {
font-family: ‘Lato’;
src: url(‘/static/fonts/Lato.woff2’) format(‘woff2’);
font-weight: 400;

@font-face {
font-family: “Lato-fallback”;
size-adjust: 97.38%;
ascent-override: 99%;
src: native(“Arial”);

h1 {
font-family: Lato, Lato-fallback, sans-serif;

Prior to those, adjusting the fallback font required utilizing the Font Loading API in JavaScript which was extra difficult, however this selection due out very quickly might lastly give us a better answer that’s extra prone to acquire traction. See my earlier article on this topic for extra particulars on this upcoming innovation and extra sources on that.

Preliminary Templates For Shopper-side Rendered Pages

Many client-side rendered pages, or Single Web page Apps, render an preliminary fundamental web page utilizing simply HTML and CSS, after which “hydrate” the template after the JavaScript downloads and executes.

It’s straightforward for these preliminary templates to get out of sync with the JavaScript model as new parts and options are added to the app within the JavaScript however not added to the preliminary HTML template which is rendered first. This then causes CLS when these parts are injected by JavaScript.

So evaluate all of your preliminary templates to make sure they’re nonetheless good preliminary placeholders. And if the preliminary template consists of empty <div>s, then use the strategies above to make sure they’re sized appropriately to attempt to keep away from any shifts.

Moreover, the preliminary div which is injected with the app ought to have a min-height to keep away from it being rendered with 0 top initially earlier than the preliminary template is even inserted.

<div id=”app” type=”min-height:900px;”></div>

So long as the min-height is bigger than most viewports, this could keep away from any CLS for the web site footer, for instance. CLS is simply measured when it’s within the viewport and so impacts the person. By default, an empty div has a top of 0px, so give it a min-height that’s nearer to what the precise top shall be when the app masses.

Guarantee Consumer Interactions Full Inside 500ms

Consumer interactions that trigger content material to shift are excluded from CLS scores. These are restricted to 500 ms after the interplay. So if you happen to click on on a button, and do some complicated processing that takes over 500 ms after which render some new content material, then your CLS rating goes to endure.

You possibly can see if the shift was excluded in Chrome DevTools through the use of the Efficiency tab to document the web page after which discovering the shifts as proven within the subsequent screenshot. Open DevTools go to the very intimidating (however very helpful when you get a dangle of it!) Efficiency tab after which click on on the document button within the high left (circled on the picture under) and work together along with your web page, and cease recording as soon as full.

You will notice a filmstrip of the web page by which I loaded a number of the feedback on one other Smashing Journal article so within the half I’ve circled, you’ll be able to nearly make out the feedback loading and the pink footer being shifted down offscreen. Additional down the Efficiency tab, underneath the Expertise line, Chrome will put a reddish-pinkish field for every shift and if you click on on that you’re going to get extra element within the Abstract tab under.

Right here you’ll be able to see that we received a huge 0.3359 rating — effectively previous the 0.1 threshold we’re aiming to be underneath, however the Cumulative rating has not included this, as a result of Had latest enter is ready to Makes use of.

Guaranteeing interactions solely shift content material inside 500 ms borders on what First Enter Delay makes an attempt to measure, however there are circumstances when the person may even see that the enter had an impact (e.g. a loading spinner is proven) so FID is nice, however the content material might not be added to the web page till after the five hundred ms restrict, so CLS is dangerous.

Ideally, the entire interplay shall be completed inside 500ms, however you are able to do some issues to put aside the mandatory area utilizing the strategies above whereas that processing is happening in order that if it does take greater than the magic 500 ms, then you definitely’ve already dealt with the shift and so is not going to be penalized for it. That is particularly helpful when fetching content material from the community which may very well be variable and outdoors your management.

Different objects to be careful for are animations that take longer than 500ms and so can affect CLS. Whereas this might sound a bit restrictive, the intention of CLS isn’t to restrict the “enjoyable”, however to set affordable expectations of person expertise and I don’t assume it’s unrealistic to count on these to take 500ms or underneath. However if you happen to disagree, or have a use case they may not have thought of, then the Chrome staff is open to suggestions on this.

Synchronous JavaScript

The ultimate approach I’m going to debate is a bit of controversial because it goes in opposition to well-known net efficiency recommendation, however it may be the one methodology in sure conditions. Principally, if in case you have content material that you recognize goes to trigger shifts, then one answer to keep away from the shifts is to not render it till it’s settled down!

The under HTML will conceal the div initially, then load some render-blocking JavaScript to populate the div, then unhide it. Because the JavaScript is render-blocking nothing under this shall be rendered (together with the second type block to unhide it) and so no shifts shall be incurred.

.cls-inducing-div {
show: none;

<div class=”cls-inducing-div”></div>


.cls-inducing-div {
show: block;

You will need to inline the CSS within the HTML with this method, so it’s utilized so as. The choice is to unhide the content material with JavaScript itself, however what I like concerning the above approach is that it nonetheless unhides the content material even when the JavaScript fails or is turned off by the browser.

This system may even be utilized with exterior JavaScript, however this can trigger extra delay than an inline script because the exterior JavaScript is requested and downloaded. That delay might be minimized by preloading the JavaScript useful resource so it’s accessible faster as soon as the parser reaches that part of code:


<hyperlink rel=”preload” href=”cls-inducing-javascript.js” as=”script”>


.cls-inducing-div {
show: none;
<div class=”cls-inducing-div”></div>
<script src=”cls-inducing-javascript.js”></script>
.cls-inducing-div {
show: block;


Now, as I say, this I’m certain will make some net efficiency folks cringe, as recommendation is to make use of async, defer or the newer sort=”module” (that are defer-ed by default) on JavaScript particularly to keep away from blocking render, whereas we’re doing the alternative right here! Nevertheless, if content material can’t be predetermined and it’ll trigger jarring shifts, then there’s little level in rendering it early.

I used this method for a cookie banner that loaded on the high of the web page and shifted content material downwards:

This required studying a cookie to see whether or not to show the cookie banner or not and, whereas that may very well be accomplished server-side, this was a static website with no potential to dynamically alter the returned HTML.

Cookie banners might be applied in numerous methods to keep away from CLS. For instance by having them on the backside of the web page, or overlaying them on high of the content material, reasonably than shifting the content material down. We most popular to maintain the content material on the high of the web page, so had to make use of this method to keep away from the shifts. There are numerous different alerts and banners that website house owners might favor to be on the high of the web page for numerous causes.

I additionally used this method on one other web page the place JavaScript strikes content material round into “principal” and “apart” columns (for causes I gained’t go into, it was not potential to assemble this correctly in HTML server-side). Once more hiding the content material, till the JavaScript had rearranged the content material, and solely then exhibiting it, prevented the CLS points that have been dragging these pages’ CLS rating down. And once more the content material is mechanically unhidden even when the JavaScript doesn’t run for some motive and the unshifted content material is proven.

Utilizing this method can affect different metrics (significantly LCP and likewise First Contentful Paint) as you might be delaying rendering, and likewise probably blocking browsers’ look forward preloader, however it’s one other instrument to think about for these circumstances the place no different possibility exists.


Cumulative Structure Shift is brought on by content material altering dimensions, or new content material being injected into the web page by late working JavaScript. On this submit, we’ve mentioned numerous suggestions and tips to keep away from this. I’m glad the highlight the Core Net Vitals have shone on this irritating situation — for too lengthy we net builders (and I positively embody myself on this) have ignored this drawback.

Cleansing up my very own web sites has led to a greater expertise for all guests. I encourage you to take a look at your CLS points too, and hopefully a few of these suggestions shall be helpful if you do. Who is aware of, you could even handle to get right down to the elusive 0 CLS rating for all of your pages!

Extra Sources

Core Net Vitals articles right here on Smashing Journal, together with my very own on Setting Width and Heights on Photographs, Measuring Core Net Vitals, and CSS Font Descriptors.
Google’s Core Net Vitals documentation together with their web page on CLS.
Extra particulars on the latest change to CLS
The CLS Changelog detailing adjustments in every model of Chrome.
The Nearly-Full Information to Cumulative Structure Shift by Jess Peck.
A Structure Shift GIF Generator to assist generate shareable demonstrations of CLS.

    About Marketing Solution Australia

    We are a digital marketing company with a focus on helping our customers achieve great results across several key areas.

    Request a free quote

    We offer professional SEO services that help websites increase their organic search score drastically in order to compete for the highest rankings even when it comes to highly competitive keywords.

    Subscribe to our newsletter!

    More from our blog

    See all posts

    Leave a Comment