Website speed optimization services, measured on a phone before and after
We measure each template the way your visitors load it, find the largest cause of delay on each one, fix it on a staging copy you approve and measure again, so the change is a number in a document.
Every site below is live and owned by the client. Each window is a capture of the real page, not a drawn mockup.
Categories we work across
HospitalityFinancial servicesManufacturing and exportConsulting and researchRecruitmentD2C
UrjalifeHospitalityAdvents WealthFinancial servicesSyounaaWellness apparelInfokoolManufacturingHoward Johnson HintonHotelSycamore ConsultancyRecruitmentStuti ExportsTechnical textilesDuhomesStudent housingVasPro GlobalConsultingPretorMarketing research
What the work covers
Three numbers that tell you why a page is slow
Website speed optimization services find what delays a page on a visitor’s phone and remove it. A single score cannot do that, because a page can be slow for four unrelated reasons. We keep three numbers for each template, and each one answers a question the score leaves open.
The field number
Field data is collected from real visits, on the phones and connections your visitors have. We read it for each template where Google reports it. Where a template has too few visits for Google to report, the review says so and leans on the lab figure instead. It is the number that matches what people experienced, and it is the one we want to move.
The lab number
A lab test loads the page on one device and one network, the same way each time. That repeatability lets us see the effect of one change at a time on a staging copy.
The cause list
We split the load time into the wait for the first byte, the delay before the main image is requested, the time it takes to download and the time to paint it. The largest part gets fixed first, so effort goes where the time is.
How it’s done
Two ways to speed up the same website
Both can shorten a load time. They differ in how much you know about why the site was slow, and in what stays fast after the next plugin or campaign tag is added.
Site-wide settingsOne layer over everything
Template by templateCause found, then fixed
Starting point
A single score for the home page, run from a desktop or a testing site
Field and lab results for the home page, a service page, a listing and a form page, each on its own
The fix
An optimization plugin or a host setting applied across the whole site
Changes to the page that is slow: the hero image, the scripts it waits for, the server’s response
Risk to the site
Switched on live, with problems found by visitors or staff
Tested on a staging copy, with forms, menus and tracking checked before release
What you can read afterwards
A list of the settings that were turned on
A change log with the measured effect of each change
The next marketing tag
No record of what a new script adds until the score drops
A script budget that says what each outside script costs in load time
When settings are enough
On a small site with one template and a few images, a plugin may be all it needs, and we will say so in the review. The template-by-template approach earns its time when the site has several page types, a page builder, a store or a stack of marketing scripts.
The facts
Six facts that shape a speed project
What Google’s web.dev, Chrome’s Lighthouse team and Google Search Central say about measuring and improving load time. Sources under each.
01
Where the time goesweb.dev
Largest Contentful Paint has four parts, and two of them should be small.
web.dev splits LCP into time to first byte, resource load delay, resource load duration and element render delay. Its guideline puts about 40% in the first byte and about 40% in load duration, and under 10% in each delay. A page far from that split points to its cause.
02
The hero imageweb.dev
Google says never to lazy-load the image that is the LCP element.
Lazy loading holds the request back until the browser confirms the image is in view, which adds load delay. web.dev advises fetchpriority “high” on the LCP image, and warns that using it on more than one or two images stops it helping.
03
What a score weighsChrome for Developers
Three metrics make up 80% of a Lighthouse 10 performance score.
Lighthouse 10 weights Total Blocking Time at 30%, Largest Contentful Paint at 25% and Cumulative Layout Shift at 25%, with First Contentful Paint and Speed Index at 10% each. A score is therefore a blend, and two sites with the same score can be slow for different reasons.
04
Lab and field differweb.dev
A lab test loads the page on a cold cache, and the field does not.
web.dev lists the reasons lab and field results differ: one device and network in the lab, a cold cache, no back/forward cache, and a page that is not personalized. Visits in the field include returning users, scrolling and varied phones. We read both and say which one a finding comes from.
05
Responsivenessweb.dev
A page can have very good Total Blocking Time and still have poor INP.
web.dev says Total Blocking Time, a lab metric, does not capture every cause of a slow response to a tap, so a page “may have poor INP despite having very good TBT scores.” Interaction delay is therefore checked in the field data, and the lab figure is read as a clue.
06
SearchGoogle Search Central
Google says its core ranking systems look at a variety of signals that align with overall page experience.
Google also says Search always seeks to show the most relevant content, even if the page experience is sub-par. Speed is one signal among several, so it supports a good page and does not replace one.
The scope
What website speed optimization produces
Each part of the work ends in something you can open: a table, a log, a setting or an account. What is covered, and what is quoted separately, is set in the written scope.
Baseline on a phone
We pick the page types that carry your traffic, such as the home page, a service or product page, a listing and the form page. Each is measured in the lab on a throttled phone profile, and read in Google’s field data where the site has enough visits. You receive: a baseline table with one row for each template.
Field dataLab testTemplates
The main image and the first paint
We find the element that counts as the largest paint on each template, then check that it is a real image tag the browser can find early, that it is sized for a phone and that nothing delays it, such as lazy loading or a slider script. Fonts that block text are handled here too. You receive: the list of changes to the main image and fonts, with the time each saved.
LCPHero imageFonts
Scripts and third parties
We list each script the page loads, who added it and what it does, then remove the ones with no job, delay the ones that can wait and load a chat or review widget when a visitor asks for it. Tracking is tested again afterward. You receive: a script inventory with the cost of each and a decision beside it.
Tag managerChatWidgets
Caching and server response
We check how long the server takes to send the first byte, whether pages are cached, whether a CDN sits in front and whether the hosting plan or PHP version is the limit. Hosting changes are proposed with the cost before anything moves. You receive: a server response record, and the cache and CDN settings written down.
Page cacheCDNHosting
Layout stability
Layout shift is traced to images and embeds without reserved space, banners that push content down and fonts that swap late. We set dimensions, reserve space and test the phone layout while the page loads. You receive: a list of the elements that moved and the change made to each.
CLSImagesEmbeds
Re-measure and handover
After release, each template is measured again with the same settings as the baseline, and the field data is read again once enough visits have been collected. We agree on a script budget for future additions. You receive: the before and after table, the change log, and the script budget your team can apply to any new tag.
BeforeAfterScript budget
How it runs
Measure, find the cause, fix on staging, measure again
You approve each stage before the next begins, and the stage hands you a table to keep.
speed-baseline.csv
Speed baseline
Sample · Sample Keystone Garage Doors, Pittsburgh, PA
Template
Pages
Field LCP (phone)
Lab LCP (throttled)
Home
1
4.1 s
4.6 s
Service page
12
3.8 s
4.2 s
Gallery
8
5.2 s
6.0 s
Blog post
24
3.3 s
3.5 s
Quote request
1
2.9 s
3.1 s
Five templates cover 46 pages: 1 + 12 + 8 + 24 + 1. Field figures are the 75th percentile for phone visits, as Google reports them.
Home page LCP breakdown
Sample · Sample Keystone Garage Doors, Pittsburgh, PA
Part
Time
Share
What sits in it
Time to first byte
1.4 s
30%
No page cache; the server builds the page for each visit
Resource load delay
0.9 s
20%
The hero is a slider background that starts loading after a script runs
Resource load duration
1.6 s
35%
The hero file is 1.1 MB for a phone-width screen
Element render delay
0.7 s
15%
A web font and a chat script hold the first paint
The parts add to 4.6 s, the lab figure in the baseline, and the shares add to 100%. Load duration and the first byte are the two largest parts, so the plan starts with the hero image, then takes the server’s first byte.
Home page change log
Sample · Sample Keystone Garage Doors, Pittsburgh, PA
Change
Lab LCP after it
Saved
Starting point
4.6 s
None
Hero set as an image tag, resized to 150 KB, fetchpriority high
3.4 s
1.2 s
Page cache switched on at the host, CDN in front
2.6 s
0.8 s
Slider script removed; chat widget loads when a visitor opens it
2.3 s
0.3 s
Forms, menu and call button tested on staging
Pass
None
The three changes save 1.2 s + 0.8 s + 0.3 s = 2.3 s, taking the home page from 4.6 s to 2.3 s in the lab. The chat widget still opens in one tap, and the tag manager still reports quote requests.
Before and after
Sample · Sample Keystone Garage Doors, Pittsburgh, PA
Template
Lab LCP before
Lab LCP after
Within the target
Home
4.6 s
2.3 s
Yes
Service page
4.2 s
2.4 s
Yes
Gallery
6.0 s
3.1 s
No
Blog post
3.5 s
2.4 s
Yes
Quote request
3.1 s
2.2 s
Yes
Four of five templates are within the target in the lab. The gallery is not: its 8 pages each carry dozens of full-size photos, so a lazy-loaded, resized gallery goes on the next list.
Our stack
The tools we use to find the delay and prove the fix
Measurement tools and the platforms where fixes are made. Pick one for the detail.
Lighthouse
Why Lighthouse
Lighthouse is Google’s open-source tester for a page’s speed, accessibility and best practices, and it runs in Chrome or from the command line.
How we use it
We run it on a throttled phone profile with the same settings each time, so a before and after compare like with like and a single change shows its own effect.
Perf auditCWVIn CI
LighthouseYes
PageSpeedNo
GTmetrixPaid
WebPageTestYes
Questions
Website speed optimization services, answered
They are the work of finding what delays a page on a visitor’s device and removing it. That means the main image, the scripts a page waits for, the server’s response and layout that shifts while loading. We measure each page type on a phone before and after, and keep the changes in a log you can read.
FAQ · 01
Speed work has no published price. It is quoted in a written scope after we measure the site, because the size of the job depends on the platform, the number of page types, how many outside scripts load and whether the host is the limit. The scope names each task, so another quote can be set beside it. A new site starts from $8,000.
FAQ · 02
Measure first, on a phone. Find the largest part of the load time, whether the first byte or the main image, and fix that before anything else. Resize and prioritize the hero image, remove scripts with no job, turn on page caching, and reserve space for images. Then measure again with the same settings so you know what each change was worth.
FAQ · 03
Slow pages lose visitors, and speed is one of the signals Google’s systems read. Google reports Core Web Vitals in Search Console, and says its core ranking systems look at a variety of signals that align with overall page experience. It also says it seeks to show the most relevant content even if the page experience is sub-par. Speed supports a good page. It does not take the place of the content a visitor came for, and we say so in the review.
FAQ · 04
Sometimes. On a small site with one template and few images, a caching and image plugin can cover the main gains. On a site with a page builder, a store or several marketing scripts, a plugin cannot tell you why a given page is slow. We test with the plugin on and off, and recommend it where it helps.
FAQ · 05
Yes. We trace load time to the theme, the page builder and each plugin that adds scripts or database queries, then change the templates, the caching and the plugin list. Changes go on a staging copy first, and the hosting and admin accounts stay in your name.
FAQ · 06
We can review one. Shopify and other hosted stores limit what we can change on the server, so the work centers on images, theme code and the apps that load scripts. We measure collection, product, cart and checkout pages separately, and say in the review which causes sit inside the platform and which are yours to change.
FAQ · 07
The test and the visit differ. A lab test loads the page on one throttled device with a cold cache, while your visitors’ phones, networks and caches vary, and many visits start from a warm cache. Lighthouse also weighs Total Blocking Time at 30%, which a site can fail without feeling slow. We read your field data beside the score before we decide what to fix.
FAQ · 08
Google’s guideline for the largest paint is 2.5 seconds or less at the 75th percentile of page loads in the field data, and Search Console groups pages by that line. We cannot promise a number, because the field result depends on your visitors’ phones, your host and your scripts. We set a target for each template in the written scope and report the measured result against it.
FAQ · 09
It can if changes are made on the live site, which is why they go on a staging copy first. We test the forms, the menu, the call button, the chat widget and the tags on staging, and you approve the set before it goes live. If a change causes a problem anyway, we revert it and log why.
FAQ · 10
You do, from day one. The host, the CDN, the tag manager and the domain are opened in your name, and we are added as users. The script inventory, the cache settings and the change log are written down, so another developer could take over without having to guess how the site was tuned.
FAQ · 11
No. Visitors’ phones, your host and the scripts your team adds later are outside our control, and Google decides how it ranks. We agree on a target for each template in your own numbers, measure before and after with the same settings, and report the result against the target, whether it was met or not.
Give us your site’s address and the platform behind it. We’ll measure your main templates on a phone, read the field data, list what delays each page and quote the work in a written scope.