Site speed
Fast enough that people stay.
Site speed is one of the few areas of marketing with published thresholds, free measurement from the platform itself, and a large public study of what improvement is worth. It is also where most sites have the most unclaimed ground.
How much does website speed affect conversions?
Measurably. Google commissioned Deloitte and 55 to study 37 brand sites across more than 30 million sessions, published in 2020, and found that a 0.1 second improvement in mobile load time was associated with an 8.4% lift in retail conversions, a 10.1% lift in travel conversions, and a 21.6% lift in progression to form submission for lead generation sites.
Those are correlations from large brands, not a promise about your site. What they establish is that speed work is conversion work rather than a technical nicety.
Google publishes the thresholds to aim at: Largest Contentful Paint under 2.5 seconds, Interaction to Next Paint under 200 milliseconds, and Cumulative Layout Shift under 0.1, measured at the 75th percentile of real page loads.
What the three numbers actually mean
Core Web Vitals sound technical and describe three ordinary experiences: how long until you can see the main thing, how long until the page responds when you tap, and how much it jumps around while loading.
Largest Contentful Paint is the first of those. It measures when the biggest visible element, usually a headline or a hero image, finishes rendering. It is the closest thing to how long until this page is useful.
Interaction to Next Paint measures responsiveness. You tap a menu and the page takes half a second to acknowledge it. That delay is usually JavaScript occupying the browser, and it is the vital most heavily plugin based sites fail.
Cumulative Layout Shift measures instability. Content moves as images and advertisements load late, and you tap the wrong thing. It is the most irritating of the three and frequently the easiest to fix.
All three are measured at the 75th percentile of real visits, which means three quarters of your actual visitors have to have a good experience. An average flattered by fast desktop users does not pass.
What a tenth of a second is worth
Commissioned by Google, conducted by 55 and Deloitte across 37 European and American brand sites, monitoring mobile load times for 30 days across more than 30 million sessions. Published in 2020.
SourceMilliseconds Make Millions, Google with Deloitte and 55, 2020, 37 sites and 30M+ sessions
State the year when you quote these, as we have. Consumer behaviour in 2020 is not identical to 2026, and the honest claim is that speed and conversion move together, not that your site will see these numbers.
How many sites actually pass
The HTTP Archive Web Almanac analyses real field data across millions of sites annually. This is an open dataset rather than a vendor study.
SourcePerformance chapter, HTTP Archive Web Almanac, 2025
Roughly half of the web fails on mobile, which is worth knowing in both directions. It is a low bar in absolute terms and clearing it still puts you ahead of half your competitors.
What actually makes a site fast
Ranked by how much difference each one usually makes on a typical business website.
Fix the images
This is almost always the largest single win and the least technical. Images uploaded at full camera resolution, served at their original size, in a format from fifteen years ago. Correct sizing, modern formats and lazy loading below the fold routinely halve page weight.
- Served at the size actually displayed, per breakpoint
- Modern formats with fallbacks
- Explicit dimensions, which also fixes layout shift
Deal with the JavaScript
Third party scripts, chat widgets, tracking tags, analytics, page builders and plugins all compete for the browser's main thread. This is what fails Interaction to Next Paint. The fix is auditing what loads, deferring what can wait, and removing what nobody uses.
Stop the layout jumping
Reserve space for images, advertisements and embeds before they load. Load fonts so that text does not reflow when the custom font arrives. This is usually a day of work and it is the most noticeable improvement for a human being.
See the remaining steps: What actually makes a site fastHide the remaining steps: What actually makes a site fast
Get hosting and caching right
Server response time sets the floor for everything else. Page caching, a content delivery network, and hosting that is not the cheapest shared plan available. On a slow host no amount of front end work reaches the thresholds.
Audit the third parties
Chat widgets, review widgets, booking embeds, heat mapping, social feeds and tag managers. Each was added for a reason and each has a cost. Loading a chat widget only on interaction rather than on page load is a common and substantial win.
Measure on the field data, repeatedly
Search Console reports Core Web Vitals from real visits to your site. That is the measurement that counts. Lab tools tell you why something is slow. Field data tells you whether it still is.
On a typical business site, images and third party scripts account for most of the available improvement, and neither requires rebuilding anything.
Why your PageSpeed score is not the target
PageSpeed Insights gives you two things and most people only look at one. The score out of 100 is a lab test: a simulated device on a simulated connection, run once. The field data section, when your site has enough traffic to have it, is what real visitors actually experienced.
A site can score 100 in the lab and fail Core Web Vitals in the field, because the lab does not have your visitors' older phones, weak signals, browser extensions or the third party scripts that behave differently in the wild. The reverse also happens.
Read the full breakdown: Why your PageSpeed score is not the targetHide the full breakdown: Why your PageSpeed score is not the target
So the score is a diagnostic. It tells you what is slow and why, which is genuinely useful when deciding what to fix. It is not the objective and it should not be the number in a report.
This matters commercially because a great deal of speed optimisation is sold against the lab score, which is the number easiest to move and least connected to what your customers experience. Ask anyone selling you speed work to show you field data before and after.
What cannot be fixed, and when to stop
There is a point past which further speed work costs more than it returns, and being told where that point is should be part of the service.
- A page builder that generates heavy markup can be mitigated and not removed without rebuilding the site.
- A booking or scheduling embed you depend on will slow the page it sits on. It can be deferred and isolated, not eliminated.
- Shared hosting on the cheapest tier sets a floor on server response time that front end work cannot get under.
- Third party review or chat widgets you have contracted for behave how they behave. Loading them later helps, removing their cost does not.
- A site on a hosted builder platform may not expose the controls needed to reach the thresholds at all.
- Very image heavy categories, such as photography portfolios, have a genuine tension between the product and the load time.
- Once you are comfortably inside the thresholds on field data, further work has little return and the budget belongs elsewhere.
The last one is the one agencies rarely say. Passing is the goal. Chasing a perfect lab score after you already pass is spending money on a number nobody sees.
How this is measured
Search Console's Core Web Vitals report, which uses field data from real Chrome users on your site, grouped by URL pattern. This is the primary measurement and it is free.
PageSpeed Insights for diagnosis when something fails, since it tells you which resource is responsible.
Read the full breakdown: How this is measuredHide the full breakdown: How this is measured
Conversion rate before and after, by device, because that is the outcome the work exists for. On a site with paid traffic, cost per lead is the cleanest signal.
Bounce and exit behaviour on the slowest pages, which usually correlate visibly enough to be worth checking.
Give field data four weeks after a change before drawing conclusions. It is a rolling 28 day window, so improvements appear gradually rather than the day the work lands, which catches people out.
We do not publish a price for this piece of work on its own, because the right scope depends on what already exists. What is published is the bundle pricing: 2,400, 3,600 or 4,800 dollars a month depending on which channels are running. You can read the full breakdown on the pricing page, and you will get an exact number in writing before anything starts.
How this connects to the rest
Speed work sits inside our web design and development service and overlaps with technical SEO, which looks at the same page from the crawling side rather than the conversion side.
It affects paid advertising directly. Google counts landing page experience inside Ad Rank, so a faster page lowers what you pay per click as well as converting more of them.
On a store it usually means auditing Shopify apps, which is where the load time went, and it is covered alongside the rest of e-commerce SEO.
Where the underlying platform cannot reach the thresholds, the conversation becomes a WordPress rebuild or a platform change rather than further optimisation.
Check your own field data first.
Open Search Console, look at the Core Web Vitals report, and see whether you actually fail. If you pass comfortably, the budget belongs somewhere else and we will tell you so.
Straight answers.
How fast should my site be?
Google publishes the thresholds: Largest Contentful Paint under 2.5 seconds, Interaction to Next Paint under 200 milliseconds, Cumulative Layout Shift under 0.1, at the 75th percentile of real page loads.
Those are the numbers to hold any provider to, and you can see them yourself in Search Console without a tool from an agency.
Is my PageSpeed score the thing to improve?
No. It is a lab simulation and a useful diagnostic. Field data from real visitors is what Google actually uses and what your customers experience.
A site can score 100 in the lab and fail in the field, which is why we report field data before and after.
Does speed affect my rankings?
It is one input among many, and a much larger factor in whether people stay and convert once they arrive.
Treat it as conversion work that also helps search, rather than as a ranking lever on its own.
Can you make any site fast?
Most sites can be substantially improved. Some cannot reach the thresholds without a rebuild, usually because of a page builder, a heavy theme, or a hosted platform that does not expose the necessary controls.
We will tell you which category you are in before taking the work rather than after.
How long does the work take, and how long until I see it?
The work itself is usually days to a couple of weeks depending on what is wrong. Seeing it in field data takes about four weeks, because the report uses a rolling 28 day window.
That lag surprises people, so it is worth knowing before the invoice arrives.
Where this comes from.
Primary documentation and published research behind the guidance on this page.
Talk to the team
A short call, a look at how the business currently shows up, and a straight answer on what we would do first.