Website performance optimisation in 2026, Core Web Vitals, INP and page speed explained
Your website is slow. Someone has told you it is hurting your Google rankings. You have been sent a PageSpeed Insights screenshot with a red score on it, and now you need to decide whether this is worth spending money on.
This guide answers that properly. It explains what Core Web Vitals are, what each metric measures in plain language, what Google changed in 2026 and what it did not, and what actually fixes a slow site.
Everything here is checked against Google's own documentation, read on 11 November 2026. Every figure is sourced and listed at the end. Where popular advice is wrong, we say so and show you the official position instead.
It is written for business owners and marketing managers in Dubai, Abu Dhabi and across the UAE, not for developers. You will not need to read any code.
The short version
If you only read one section, read this one.
- There are three Core Web Vitals, and they have not changed. Largest Contentful Paint, Interaction to Next Paint and Cumulative Layout Shift. The thresholds are 2.5 seconds, 200 milliseconds and 0.1.
- Google did not add a new metric in 2026. If you have read that it did, that article was wrong. Google's own documentation still lists three stable metrics.
- Something real did change in October 2026. Google started publishing experimental versions of all three metrics that include single-page app navigations. Google says it aims to make them official in May 2027, and that Search Console will then switch to the new data.
- Only about 56 per cent of websites pass all three. That is Google's own figure. Failing is normal, which also means passing is a genuine advantage.
- Interaction to Next Paint is getting worse across the web, and Google has said publicly that it does not know why.
- Your PageSpeed Insights score is not your Core Web Vitals. This is the single most common misunderstanding, and it wastes a lot of money. The two are measured completely differently.
- Google measures real visitors, not your laptop. It uses the 75th percentile of actual page loads, split between mobile and desktop.
- Speed is a ranking factor, but a modest one. Google confirms Core Web Vitals are used by its ranking systems, and separately says chasing a perfect score purely for SEO may not be the best use of your time.
- Most problems come from four things. Too much JavaScript, images that load late, elements with no reserved space, and a slow server.
- This is development work, not marketing work. Anyone who promises to fix it from outside your codebase is selling you something that does not exist.
What this means for you: a slow site is worth fixing because visitors leave and because it is one of the ranking inputs you can fully control. It is not worth fixing to turn a number green.
Find out what is actually slowing your website down
Tomsher runs website performance audits for UAE businesses. We look at your real visitor data rather than one test run, identify which of the three metrics you are failing and why, and give you a prioritised list with the development effort beside each item. Then you decide what is worth doing.
Tomsher Technologies, operating in Dubai since 2011 with a fully in-house development team. Rated 4.9 out of 5 from 201 Google reviews.
What Core Web Vitals are, in plain language
Core Web Vitals are three measurements Google takes from real people using your website. Each one covers a different part of the experience.
Here they are with their thresholds, taken from Google's documentation.
| Metric | What it measures | Good | Needs work | Poor |
|---|---|---|---|---|
| Largest Contentful Paint, LCP | Loading. How long until the biggest thing on screen appears | 2.5 seconds or less | 2.5 to 4 seconds | Over 4 seconds |
| Interaction to Next Paint, INP | Responsiveness. How long the page takes to react when someone taps or clicks | 200 milliseconds or less | 200 to 500 ms | Over 500 ms |
| Cumulative Layout Shift, CLS | Stability. How much the page jumps around while it loads | 0.1 or less | 0.1 to 0.25 | Over 0.25 |
Largest Contentful Paint, in one paragraph
Someone opens your page. Something big appears. Usually that is your hero image, a product photo or a block of heading text. LCP is the time from the page starting to load until that element is painted on screen. It is a proxy for the question a visitor is actually asking, which is whether anything is happening yet.
Google says 40 per cent of sites in its dataset fail this one. It is the metric businesses struggle with most.
Interaction to Next Paint, in one paragraph
Someone taps a button, opens a menu, or types in a field. INP measures how long it takes before the screen visibly updates in response. It is not about loading at all. It is about whether the page feels dead when you touch it.
INP replaced an older metric called First Input Delay in March 2024. The difference matters: First Input Delay only measured the delay before the browser started working, and only on the first interaction. INP measures the full time until something appears on screen, and it considers interactions throughout the visit. It is a much harder test to pass, which is the point.
Cumulative Layout Shift, in one paragraph
You start reading a paragraph, an advert loads above it, and the text jumps down. You go to tap a button and something shifts under your finger. CLS is a score, not a time, representing how much of that happened. Lower is better.
About a quarter of sites still fail this, and it is usually the easiest of the three to fix.
The two words that explain most confusion
Field and lab. You need both terms to understand anything else in this article.
- Field data is collected from real people using your site in real conditions. Google collects it through Chrome and publishes it in something called the Chrome User Experience Report. This is what counts for Core Web Vitals and for search.
- Lab data is a single simulated test run on a server somewhere, under controlled conditions. Lighthouse and the performance score in PageSpeed Insights are lab data.
Lab data is useful for diagnosis. It is not your Core Web Vitals. More on why that matters below, because it is where most of the wasted effort happens.
Core Web Vitals, explained without the jargon
Not sure which of the three you are failing
It is the first question worth answering, because the three metrics have completely different causes and completely different fixes. We will check your real visitor data and tell you, at no charge, before anyone quotes you for anything.
What Google did not change in 2026
Worth stating clearly, because a great deal of published advice about 2026 is simply invented.
Search for Core Web Vitals updates and you will find articles announcing a new fourth metric, a composite score, revised thresholds, or a change that makes page speed the deciding ranking factor. We checked Google's own documentation on 11 November 2026. None of that is there.
As of today:
- There are three stable Core Web Vitals, and they are LCP, INP and CLS.
- The thresholds are still 2.5 seconds, 200 milliseconds and 0.1.
- They are still assessed at the 75th percentile, split by mobile and desktop.
- There is no composite Core Web Vitals score.
- Google states that stable Core Web Vitals will not change more than once per year, and that any change will be announced in advance with a documented changelog.
What this means for you: if an agency pitches you urgent work because of a 2026 Core Web Vitals update, ask them which Google page says so. The thresholds that applied in 2024 still apply now.
What Google actually changed, and why May 2027 is the date to note
There is a real change, and it is bigger than anything the invented versions describe. It just affects a specific kind of website.
The problem Google is solving
Most websites load a fresh page from the server every time you click a link. Those are called hard navigations.
Some websites do not. A single-page application, often shortened to SPA, loads once and then swaps the content and the address bar using JavaScript without a full reload. React, Vue, Angular and Next.js sites often work this way. Those in-app transitions are called soft navigations.
Until October 2026, Google measured only the first hard load of such a site. Everything after that was invisible as a separate page view, and all the interactions and layout shifts from the entire session were lumped onto that one initial page. The result was a distorted picture, and it made SPAs look worse than they were.
What Google launched
Following the arrival of a Soft Navigations browser feature in Chrome 151, Google started publishing three experimental metrics in the October 2026 Chrome User Experience Report release. They are versions of LCP, INP and CLS that include soft navigations, and they are available in Google's BigQuery dataset and its public data interfaces.
Two things about them matter commercially.
First, each soft navigation is now measured as its own page view, and reported under the address the user moved to rather than the page they arrived on. For anyone trying to work out which part of an app is slow, that is a large improvement.
Second, and this is the date to write down: Google says it aims to graduate these metrics out of experimental status in May 2027, and that once they do, Search Console will switch to using the new data.
Who this affects, with Google's own numbers
Most businesses, including most in the UAE, will see no change at all. Google reports that between 66 and 75 per cent of websites have fewer than five per cent soft navigations, and that only around 7 per cent on Android and 11 per cent on desktop exceed half.
But among the top 1,000 websites, 35 per cent on Android and 42 per cent on desktop are above that halfway mark. In other words, this change is concentrated in large, app-like sites.
For those sites, the effect is substantial and positive. Google reports that among websites where 75 to 100 per cent of page views are soft navigations, the share with good LCP on Android moves from 26 per cent to 87 per cent once soft navigations are counted. INP pass rates roughly double. CLS improves too.
Across the whole web the difference is smaller but still real. In Google's September 2026 dataset, 55.9 per cent of websites had good Core Web Vitals. Counting soft navigations, that figure would be 59.5 per cent.
What this means for you: if your site is a single-page application, your Search Console numbers are likely to improve in 2027 without you doing anything, because you were being measured unfairly. If it is a normal multi-page website, nothing changes. Either way, do not let anyone sell you a fix for this. It is a measurement correction, not a problem with your site.
One caveat from Google worth passing on. Even with the improvement, app-heavy sites still trail ordinary multi-page sites on good INP. Heavy JavaScript work during page transitions remains a real problem, and counting it more fairly does not make it go away. If you are choosing a technology stack, our rundown of the top web development frameworks in 2026 covers the trade-offs, and ecommerce web application architecture goes deeper on the structural decisions.
Running a React or Next.js site and unsure how it will be measured. Ask our developers
The INP problem nobody has explained
Here is something you will not read in most performance guides, because it requires reading Google's monthly release notes rather than its marketing pages.
Interaction to Next Paint is getting worse across the web, and Google has said so without being able to explain it.
In Google's August 2026 release notes, the Chrome team wrote that the continued regression of INP in particular is a cause for concern, and that they do not have a definitive reason for it.
The numbers behind that statement:
| Dataset | Websites with good INP |
|---|---|
| March 2026 | 87.2 per cent |
| April 2026 | 87.1 per cent |
| May 2026 | 86.6 per cent |
| June 2026 | 85.9 per cent |
| July 2026 | 85.7 per cent |
| August 2026 | 85.3 per cent |
| September 2026 | 85.4 per cent |
That is a slide of almost two percentage points across the whole web in six months, which at that scale is millions of websites getting less responsive.
What this means for you: two things. Responsiveness is drifting in the wrong direction industry-wide, most likely because sites keep adding JavaScript, so standing still means slowly falling behind. And if your own INP has worsened without anything obvious changing, you are not imagining it and you are not alone. Check what has been added to your tag manager in the last year before you look anywhere else.
Why your PageSpeed Insights score is not your Core Web Vitals
This section exists because this one misunderstanding causes more wasted money than any other topic in technical SEO.
What the score is
When you run a page through PageSpeed Insights you get a number out of 100, usually in a coloured circle. That number is a Lighthouse performance score. It comes from a single simulated test, on Google's hardware, on a throttled connection, with no real person involved.
It is a weighted blend of lab measurements. It is useful for finding problems. It is not what Google uses to judge your page experience.
What Google actually uses
Google uses field data from real Chrome users. Specifically, it looks at the 75th percentile of your page loads, separately for mobile and desktop.
The 75th percentile matters more than people realise. It means that if three quarters of your visitors get an LCP under 2.5 seconds, you pass, even if the slowest quarter are having a worse time. It also means a handful of very fast loads cannot rescue you. You are judged by something close to your typical bad visit, not your average and not your best.
Why the two disagree so often
Four reasons, and all four are common:
- Lighthouse cannot measure INP at all. There is no user in a lab test, so there are no interactions to measure. Google's documentation confirms this and says to use Total Blocking Time as a rough lab stand-in instead.
- Real visitors have older phones and worse connections than Google's simulated test, or sometimes better ones. Either way the simulation is not them.
- Real visitors arrive with warm caches, from the back button, or from a browser that restored the page from memory. A lab test is always a cold first load.
- Layout shifts caused by cookie banners, chat widgets and adverts often appear only when a real person is there to trigger them.
What this means for you: use the score to find things to fix. Use the field data to decide whether you have a problem at all. If your field data is green and your score is amber, you do not have an emergency, whatever the screenshot looks like.
One more trap worth knowing
Google's field data comes with a lower limit. A page or a website needs enough real traffic before Google will report data for it. If your site is small or the page is new, your field data may simply be missing, and a blank figure is not the same as a bad one.
Google says the same thing about its other reporting: a blank metric can mean not enough data yet, not zero activity.
Why UAE businesses misjudge their own website speed
A pattern we see repeatedly with Dubai and Abu Dhabi clients, and it is worth naming because it is structural rather than careless.
The office test is the wrong test
Your team checks the website on a current iPhone or a recent Samsung, on fast office internet, in a city with strong mobile coverage. The site feels fine. It is fine, on that device, on that connection.
Your customers are a much wider group. Some are on three year old mid-range Android phones. Some are on a hotel connection, in a basement car park, on a bus, or roaming. Google's measurement includes all of them, weighted towards the slower end by the 75th percentile rule.
A fast device and a fast connection hide two of the most expensive problems on a website: too much JavaScript, which a powerful processor absorbs without complaint, and a slow server response, which a fast connection partially masks. On a mid-range phone, neither is hidden at all.
Where your server is matters
If your audience is in the UAE and your hosting is in Europe or North America, every single request pays a round trip across the world before anything appears. That is Time to First Byte, and it is the foundation everything else is built on.
Google reports that only 33 per cent of HTML document requests are served from a content delivery network, which is a large missed opportunity given how much of the problem it solves. A content delivery network keeps copies of your site close to your visitors, so a customer in Dubai is served from nearby rather than from Frankfurt.
This is usually the lowest-cost meaningful improvement available to a UAE website, and it is the one most often skipped. We cover the detail on our CDN integration page, and the hosting decision behind it in VPS versus cloud hosting and cloud hosting in 2026.
Arabic and bilingual sites have an extra consideration
If you run an Arabic version of your site, you are usually loading additional fonts for Arabic script. Web fonts are a frequent cause of both slow text rendering and layout shift, because text often appears in a fallback font first and then moves when the real font arrives.
It is fixable, and it needs doing deliberately rather than left to whatever the theme does by default.
Website speed optimisation UAE
Test your site the way your customers experience it
We test on mid-range devices and realistic connections, from the markets you actually sell in, and we compare that against your real visitor data. It usually produces a different list from the one your developer got off a desktop test.
How to fix Largest Contentful Paint
LCP is the metric most sites fail, so it is usually where the work starts. Google's own recommendations fall into three groups, and the figures below are Google's.
One, make the main image easy for the browser to find
According to the 2024 Web Almanac, 73 per cent of mobile pages have an image as their largest element. So in most cases, fixing LCP means fixing one image.
The counterintuitive part is that the image is rarely slow to download. Google's analysis of real Chrome data found that most websites with poor LCP spend under 10 per cent of that time actually downloading the image. The time goes somewhere else entirely.
Where it goes is waiting. On pages with poor LCP, the main image starts loading 1,290 milliseconds late at the 75th percentile, which is more than half the entire budget for a good score. The browser simply did not know the image existed yet.
Google found that 35 per cent of pages hide their main image from the browser's early scan, and 7 per cent hide it behind a JavaScript-dependent attribute that guarantees a delay. Only 15 per cent of eligible pages use the one attribute that tells the browser to prioritise it.
What this means for you: ask your developer one question. Is the main image on our key landing pages present in the HTML the server sends, and is it marked as high priority. If the answer involves JavaScript loading it in afterwards, that is your LCP problem and it is fixable.
Two, serve the page from somewhere near your visitors
Covered above. A content delivery network, and caching your pages even briefly, addresses the time before anything can begin.
Three, make the next page instant
The fastest possible page load is one that already happened. Browsers support two mechanisms for this.
The back and forward cache restores a page the visitor has already seen from memory, instantly and exactly as they left it. Google says its arrival in 2022 produced the biggest single improvement in layout stability it recorded that year. Many sites accidentally disable it through caching headers or old event handlers, and the fix is usually small.
Speculative loading lets a site load the next page before the visitor clicks, when it can predict where they are going. Used carefully it produces near-instant navigation. Used carelessly it wastes your server capacity and your visitors' data, so it needs to be driven by your actual analytics rather than guesswork.
Most LCP problems are one image on one template. Have us find yours
How to fix Interaction to Next Paint
INP is almost always a JavaScript problem. Google's recommendations come down to three ideas.
One, stop blocking the browser for long stretches
A browser does one thing at a time on its main thread. Any piece of work that takes more than 50 milliseconds is called a long task, and while it runs, your page cannot respond to a tap.
The fix is to break long work into smaller pieces and hand control back to the browser between them, so an interaction can be served in the gaps. Modern browsers provide a dedicated mechanism for exactly this.
Two, ship less JavaScript
Google's position is blunt. Websites are shipping more JavaScript than ever and the trend is not changing, and the result is competing demands on a single thread. This is almost certainly connected to the industry-wide INP slide described earlier.
In practice, for a typical UAE business website, the biggest offenders are:
- Tracking and marketing tags accumulated over years in a tag manager, many for tools nobody uses now.
- Chat widgets, booking widgets, review widgets and social feeds, each loading its own framework.
- Plugins. A WordPress site with forty plugins is loading forty sets of scripts, most on every page.
- Sliders and animation libraries used for one element on the homepage.
- Code shipped for features further down the page that nobody has scrolled to yet.
Google specifically recommends auditing your tag manager periodically and removing old tags, which is one of the few items on this list a marketing team can own directly.
What this means for you: ask who last reviewed your tag manager and your plugin list. If the answer is nobody, you have found free performance.
Three, keep the page itself simple
Very large and complex pages are expensive for the browser to recalculate every time something changes. Google's recommendations here include keeping the page structure small, avoiding patterns that force the browser to recalculate layout repeatedly, and using a modern CSS feature that lets the browser skip work for content that is off screen.
This is genuine development work, and it is where a development team earns its fee rather than a plugin. It also overlaps with design decisions, since responsive web design done badly produces pages that are heavy on every screen size rather than appropriate to each one.
INP optimisation, the hardest of the three
JavaScript reduction needs developers, not plugins
No caching plugin fixes Interaction to Next Paint. It takes someone reading your code, finding what blocks the browser, and restructuring it. That is an in-house development team, which is what Tomsher has had in Dubai since 2011.
How to fix Cumulative Layout Shift
The good news. This is normally the quickest of the three to fix, and the least costly.
One, reserve the space before the content arrives
Layout shift happens because something arrives late and pushes everything else out of the way. If the space was reserved in advance, nothing moves.
Google reports that 66 per cent of pages have at least one image with no declared dimensions. Without them, an image occupies no height until it loads, and then suddenly takes up a few hundred pixels. That is a layout shift, and setting width and height fixes it.
For things whose size you cannot know in advance, such as adverts or embeds, modern CSS provides a way to declare a shape, and a sensible minimum height is almost always better than letting a container start at zero.
Two, animate the right properties
This one surprises people. Google's data shows that pages animating any CSS property capable of affecting layout are 15 per cent less likely to have good CLS than pages overall, and that pages animating margins or border widths are rated poor at almost twice the normal rate.
Cookie banners and notification bars that slide in are a frequent culprit. Animated the wrong way they contribute to your score. Animated the right way, using properties the browser can handle without recalculating layout, they contribute nothing.
The rule Google gives is simple: never animate a property that forces the browser to redo the page layout, unless it happens in direct response to a tap or key press.
Three, make sure the back button is instant
The back and forward cache mentioned earlier eliminates layout shift entirely on a restored page, because the page comes back fully formed. Making your pages eligible for it is one change that improves both LCP and CLS at once.
The UAE-specific version of this problem
Three things cause most of the layout shift we find on UAE business websites. Cookie consent banners, which arrived with privacy compliance work and were rarely tested for stability. Arabic and English font swapping. And third-party embeds for maps, booking systems and payment options that reserve no space.
None of those are difficult. They are just nobody's job until someone makes them somebody's job.
Want the quick wins done first, in a week rather than a quarter. Website maintenance and AMC
The order to do website performance optimisation work in
Page speed work goes wrong when it is done as a list of tickets rather than in sequence. This is the order we use on website performance optimisation projects, and the sequence matters more than any individual item on it.
- Get real visitor data first. Before touching anything, find out which metric you are actually failing, on which device type, on which pages. Optimising LCP when your problem is INP is common and completely wasted.
- Fix the foundation. Server response time, caching and a content delivery network. Everything else sits on top of this, and improvements elsewhere are partly masked until it is right.
- Fix the main image on your most important pages. Highest return per hour of work on most websites.
- Remove what you are not using. Old tags, dead plugins, unused scripts, abandoned tracking tools. This is the only step that makes the site faster while also making it simpler.
- Reserve space for everything that loads late. Images, banners, embeds. Low cost, quick, visible.
- Then do the harder JavaScript work. Breaking up long tasks, splitting code, reducing what loads before the page is usable. This is where the remaining INP gains are, and it is the most expensive step, which is why it comes last.
- Measure again after four weeks. Field data is collected over a rolling window, so it does not update the moment you deploy. Patience is part of the process.
- Then protect it. Performance decays. Every new plugin, tag, embed and campaign script takes a little back.
That last point is the one businesses underestimate. We have seen sites optimised to a good state and back in the red within six months, not because anyone did anything wrong, but because nobody owned it. Ongoing website maintenance is what stops that, and it costs far less than doing the optimisation twice.
Which tools to use for what
There are a lot of tools and they do not all do the same job. This is what each is actually for.
| Tool | Data type | Use it for |
|---|---|---|
| Search Console Core Web Vitals report | Field | Your official standing with Google, grouped by page type. Start here |
| PageSpeed Insights | Both | Field data at the top, lab diagnosis below. Read both halves, not the circle |
| Chrome User Experience Report | Field | Your own and your competitors' real-world data, month by month |
| Chrome DevTools | Lab | Finding the cause of a specific problem. Developer tool |
| Lighthouse | Lab | Catching regressions before release. Cannot measure INP |
| Your own real user monitoring | Field | Per-page detail and fast regression alerts. Google recommends every site set this up |
Two notes on this table. Google strongly recommends that sites set up their own real user monitoring rather than relying on its public dataset alone, because its data is aggregated and does not give you the per-page detail you need to diagnose a problem quickly. And the Looker Studio dashboard many agencies used for this was retired by Google at the end of November 2025, so if your reporting stopped updating around then, that is why.
Our rundown of the best SEO tools in 2026 covers the wider set.
Want this owned on a schedule rather than fixed once. Website AMC from Tomsher
Does page speed actually affect your Google rankings
Yes, and less than you have been told. Both halves of that are Google's own position, so here they are.
Google states that Core Web Vitals are used by its ranking systems, and recommends that site owners achieve good Core Web Vitals for success in Search.
Google also states, on the same page, that good results in Search Console or third-party tools do not guarantee your pages will rank at the top, that there is more to page experience than Core Web Vitals scores alone, and that trying to get a perfect score purely for SEO reasons may not be the best use of your time.
It adds two points that are useful to know:
- There is no single page experience signal. Google's ranking systems look at a variety of signals.
- Google Search will still show the most relevant content even when the page experience is poor. But where many pages are equally helpful, a good experience can contribute.
What this means for you: speed is a tie-breaker, not a trump card. It will not make a thin page outrank a better one. In a competitive UAE market where several agencies, clinics or suppliers have genuinely comparable pages, it is one of the few inputs entirely within your control. That is the honest commercial case, and it is a good enough one.
The stronger argument for website performance optimisation has nothing to do with ranking. People leave slow websites. Your enquiry form converts worse when the page stutters, which is also the point of the relationship between SEO and user experience. That shows up in revenue whether or not it shows up in rankings, which is the point we make in our guide to traffic that is not turning into leads.
What slow page speed costs a UAE business
Not a list of statistics, because the credible ones are not about your market. A list of the specific consequences of poor website performance that we see in UAE accounts.
| Where it shows up | What actually happens |
|---|---|
| Paid search and social | You pay for the click, then lose the visitor before the page renders. Every slow page makes your media budget less efficient |
| Mobile enquiries | Forms that lag while being filled in get abandoned. This is an INP problem that never appears in a desktop test |
| Ecommerce checkout | Layout shift and unresponsive buttons at the payment step, which is the most expensive place on the site to lose someone |
| Organic rankings | A modest handicap against comparable competitors, in exactly the categories where competitors are comparable |
| AI search visibility | Pages that are slow or render only through JavaScript are harder for any automated system to retrieve and quote |
| Internal credibility | Nobody wants to send a prospect to a website that feels broken, so the sales team stops using it |
That fifth row is becoming the most interesting one. The technical qualities that make a page fast for people also make it readable by AI assistants, which is a point we develop in our work on generative engine optimisation in Dubai.
Send us your website and we will tell you which metric is costing you
We will look at your real visitor data, test on realistic devices, and come back with the three or four changes that would move you from failing to passing, with the development effort beside each one. If the answer is that your site is fine and the screenshot was misleading, we will tell you that instead.
M05, Wafi Residence, Oud Metha Road, Umm Hurair 2, Dubai. Rated 4.9 out of 5 from 201 Google reviews.
Running an online store where speed has a direct price attached. Ecommerce development Dubai
When optimising is the wrong answer
Said plainly, because we would rather tell you this before you spend money than after.
Sometimes a website cannot be made fast. If it is built on an old theme with a page builder, carrying years of plugins, on shared hosting, with a database nobody has maintained, then performance work becomes an expensive series of patches on a structure that will keep producing the same problems.
The signs that you are in that position:
- Every fix creates a different problem somewhere else.
- Nobody can safely remove anything because nobody knows what depends on what.
- The site slows down again within weeks of every improvement.
- Your developers quote in weeks for changes that should take hours.
At that point a website rebuild costs less over two years than maintaining the current one, and you get a faster site as a by-product rather than as a project. If you go that route, migrate it properly and work through a redesign SEO checklist first, because a rebuild that loses your rankings is not a saving.
We will tell you which of the two situations you are in. It is not in our interest to sell you a rebuild you do not need, and it is not in your interest to buy optimisation work that cannot succeed.
How Tomsher does website performance optimisation
Tomsher has operated in Dubai since 2011, with more than fifteen years of delivery experience and a fully in-house team of developers, designers and SEO specialists. We are rated 4.9 out of 5 from 201 Google reviews. Performance work sits with the people who write the code, which is the only place it can actually be done.
| What we do | What it changes |
|---|---|
| Technical SEO and Core Web Vitals work | Diagnosis from real visitor data, then the fixes, in priority order |
| Performance and SEO audits | What is wrong, what it is worth, and what it takes. Before any commitment |
| CDN integration | Your pages served close to your customers instead of across the world |
| Cloud hosting and VPS hosting | The foundation. No amount of front-end work fixes a slow server |
| Web development and website development in Dubai and Abu Dhabi | Sites built server-rendered and light from the start, rather than optimised later |
| WordPress development and WordPress amendments | Plugin reduction, theme work and the cleanup most WordPress sites need |
| Shopify, WooCommerce and Magento development | Store performance, where every lost second has a price attached |
| Ecommerce website development | Catalogue, images and checkout built to stay fast as the store grows |
| Website redesign and migration | When the existing site cannot be made fast, done without losing rankings |
| Maintenance and AMC | Keeping the gains. Performance decays, and someone has to own it |
What we will not promise
- We will not promise a score of 100. It is achievable on a brochure page and not on a real commercial site with tracking, chat and payment integrations. Anyone promising it is either removing things you need or testing the wrong way.
- We will not promise a ranking improvement from speed work. Google calls Core Web Vitals one of many signals and says a good score guarantees nothing. We agree with Google.
- We will not promise a conversion rate uplift percentage. There is no honest figure for your site that is not measured on your site.
- We will not quote performance work without looking at your real visitor data first. A quote based on one lab test is a guess.
- We will not tell you the 2026 Core Web Vitals update means you must act now. There was no such update, and we are not going to invent urgency that Google's documentation contradicts.
Frequently asked questions
What are the Core Web Vitals thresholds in 2026
Largest Contentful Paint should be 2.5 seconds or less, Interaction to Next Paint should be 200 milliseconds or less, and Cumulative Layout Shift should be 0.1 or less. Each is assessed at the 75th percentile of real page loads, measured separately for mobile and desktop, and a page is considered to pass only when all three meet their target. These thresholds have not changed in 2026 and Google states that stable Core Web Vitals will not change more than once a year.
Did Google add a new Core Web Vital or change the thresholds in 2026
No. Google's own documentation, checked on 11 November 2026, still lists three stable metrics with the same thresholds and no composite score. What Google did do, in October 2026, is start publishing experimental versions of all three metrics that include single-page app navigations. Those sit alongside the stable metrics and do not replace them. Articles describing a new fourth metric or a new scoring system in 2026 are not supported by anything Google has published.
What is INP and how is it different from First Input Delay
Interaction to Next Paint measures how long a page takes to visibly respond when someone taps, clicks or types, and it looks at interactions throughout the visit. First Input Delay, which it replaced in March 2024, measured only the delay before the browser began processing the very first interaction, and not the time until something actually appeared on screen. INP is a considerably harder test, which is why many sites that comfortably passed First Input Delay struggle with it.
Why is my PageSpeed Insights score different from my Core Web Vitals
Because they measure different things. The score out of 100 comes from a single simulated lab test on Google's hardware with no real user involved. Your Core Web Vitals come from real Chrome visitors, at the 75th percentile, over a rolling window. Lighthouse cannot measure INP at all, since there is no user in a lab test to interact with the page. Use the score to find problems to investigate and use the field data to decide whether you have a problem at all.
What is a good page speed for a website in the UAE
The same as anywhere, because Google applies one global standard: a Largest Contentful Paint of 2.5 seconds or less for at least 75 per cent of your visitors. The UAE-specific part is not the target but how you reach it. If your hosting sits in Europe or North America while your customers are in Dubai, every request pays an intercontinental round trip before anything appears, which is why a content delivery network and sensibly located hosting usually give UAE businesses their largest single improvement.
How long does it take to fix Core Web Vitals
The fixes themselves range from hours to weeks depending on which metric is failing. Reserving space for images and removing unused tracking tags can be done in days. Reducing JavaScript enough to change INP is a development project. Separately, you then wait for the measurement to catch up, because Google's field data is collected over a rolling window and will not reflect a deployment immediately. Allow around four weeks after the work before judging the result.
Is page speed a Google ranking factor
Yes, but a modest one. Google states that Core Web Vitals are used by its ranking systems and recommends achieving good scores. Google also states that good scores do not guarantee high rankings, that there is no single page experience signal, that it will still show the most relevant content even when the experience is poor, and that chasing a perfect score purely for SEO reasons may not be the best use of your time. The practical reading is that speed is a tie-breaker between comparable pages rather than something that overcomes weaker content.
What is the most common cause of a failing LCP
Waiting rather than downloading. Google's analysis found that most sites with poor Largest Contentful Paint spend under 10 per cent of that time actually downloading the main image, while on poor-LCP pages the image starts loading 1,290 milliseconds late at the 75th percentile. The usual cause is that the image is not visible in the HTML the server sends, so the browser cannot discover it early. Google found 35 per cent of pages hide their main image from the browser's early scan.
Why does my website score well on desktop but badly on mobile
Because mobile and desktop are assessed separately, and mobile visitors have slower processors and less reliable connections. A powerful laptop absorbs heavy JavaScript without appearing slow, and a fast office connection partly masks a slow server. A mid-range phone does neither. This is also why testing on a current flagship handset in a Dubai office gives a misleading result for most of your audience.
How many websites actually pass Core Web Vitals
About 56 per cent. In Google's September 2026 dataset, covering over 18 million websites, 55.9 per cent had good Core Web Vitals overall, with 68.3 per cent passing LCP, 81.7 per cent passing CLS and 85.4 per cent passing INP. That means failing is unremarkable, and it also means passing is a real differentiator, particularly against local competitors who have never looked at it.
What are soft navigation metrics and do they affect my website
They are experimental versions of the three Core Web Vitals that count in-app page changes on single-page applications, which Google previously ignored. If your site loads a fresh page from the server on every click, which describes most business websites, nothing changes for you. If your site is built as a single-page application in a framework such as React, Vue or Next.js and visitors move around within it, your measured scores are likely to improve, because you were previously having an entire session attributed to one page load. Google aims to graduate these metrics in May 2027, at which point Search Console will switch to the new data.
Can a WordPress website pass Core Web Vitals
Yes, and plenty do. WordPress itself is not the problem. The problems are usually an accumulation of plugins each loading their own scripts, a heavy page-builder theme, images without declared dimensions, and shared hosting. Removing what is unused, reducing what loads before the page is usable, and moving to appropriate hosting with a content delivery network resolves most WordPress performance complaints without changing platform. Our WordPress SEO services in Dubai cover this alongside the rest of the technical work.
Should I rebuild my website or optimise it
Optimise first, unless the signs point the other way. The signs that a rebuild is the better value are that every fix breaks something else, nobody can safely remove anything because nobody knows what depends on what, the site slows down again within weeks of each improvement, and simple changes are quoted in weeks. In that position, performance work becomes a series of patches on a structure that keeps producing the same problems, and a rebuild costs less over two years. We will tell you which situation you are in before you commit.
How do I stop my website getting slow again
By making it somebody's job. Performance decays through normal activity, since every new plugin, tracking tag, chat widget, embed and campaign script takes a little back. The practical approach is a monthly check of your field data, a review of your tag manager and plugin list on a schedule, a lab test before each release to catch regressions, and your own real user monitoring so you find out about a problem before your rankings do. This is what a maintenance contract should cover, and it costs far less than paying for the same optimisation twice.
Sources
Every metric, threshold and figure in this article comes from one of the following, all read in full on 11 November 2026. Where Google's position is stated, it is Google's wording rather than our summary of it.
- web.dev, Web Vitals. The definition of the current stable set, the thresholds, the 75th percentile rule, the metric lifecycle, and the confirmation that Lighthouse cannot measure INP. web.dev
- web.dev, Experimental soft navigation metrics now available in CrUX, by Annie Sullivan, 13 October 2026. The source for the soft navigation change, the share of websites affected, the metric-by-metric impact, and the statement that Google aims to graduate the metrics in May 2027 and that Search Console will then switch to the new data. web.dev
- Chrome UX Report release notes. The source for every pass-rate figure, for the September 2026 dataset covering 18,376,044 websites, and for the Chrome team's August 2026 statement that the continued regression of INP is a cause for concern without a definitive reason. Also records the retirement of the CrUX Dashboard at the end of November 2025. developer.chrome.com
- web.dev, The most effective ways to improve Core Web Vitals. The source for Google's own priority recommendations and for every optimisation statistic quoted here, including the 1,290 millisecond image delay, the 35 per cent of pages hiding their main image, the 66 per cent with unsized images, the 33 per cent of HTML served from a CDN, and the animation findings. web.dev
- Google Search Central, Understanding page experience in Google Search results, last updated 22 September 2026. The source for both halves of the ranking question, including that Core Web Vitals are used by Google's ranking systems and that chasing a perfect score purely for SEO may not be the best use of your time. developers.google.com
- Google Search Central, Understanding Core Web Vitals and Google Search results. Google's search-side statement of the same three thresholds. developers.google.com
- The web-vitals library changelog on GitHub. For the measurement library's own history, including soft navigation support arriving in version 6 in July 2026 and the removal of the old First Input Delay function in May 2025. github.com
Three things are deliberately absent. We quote no conversion or revenue uplift percentage for a speed improvement, because no credible figure exists for your specific website that was not measured on your website. We quote no UAE network speed figure, because we could not verify one against a primary source. And we report no October 2026 pass rates, because those release notes had not been published when this article was written, so the September 2026 dataset is cited as the most recent available. Google publishes the dataset on the second Tuesday of each month.
Website performance and development services from Tomsher
The work behind a fast website, and the people who do it.
By Digital Team. Updated on 11-11-2026
website performance optimisation website speed optimisation UAE page speed optimisation Dubai
16-11-2026
Best Ecommerce consulting services in Dubai
14-11-2026






