A website that loads fast on a phone in Iraq
Why a site that is quick in the office is slow on customers' phones, how to measure it, what to fix first, and how to write speed into a contract.
Your website opens at once on the office computer. Then a customer writes on WhatsApp that the page "keeps loading", and someone who tapped your Instagram link gives up before the first photo appears. Both of you are right. You are testing one website and your customers are visiting another.
Why the office test misleads
Three things differ between your desk and a customer's hand.
- The connection. You are on office Wi-Fi over a fixed line. The customer is on mobile data, where the signal changes from one street to the next and weakens inside buildings.
- The device. A laptop works through a heavy page far faster than a mid-range phone does. The phone has to download the page and then build it, and both steps take it longer.
- The cache. Your browser saved the logo, the fonts and the images on earlier visits, so it fetches almost nothing. A new customer downloads everything.
There is also the cost. It is common in Iraq to buy mobile data in bundles, so a page that pulls many megabytes spends the visitor's money as well as their time.
How to see what your customers see
Two free checks take about ten minutes together.
Run PageSpeed Insights. This is a free tool from Google. Enter an address and choose the mobile tab instead of desktop. Google judges a page on three measures it calls Core Web Vitals:
| Measure | What the visitor notices | Google's "good" level | When the tool shows it |
|---|---|---|---|
| LCP | How soon the main content of the page is visible | Within 2.5 seconds | Always, in the simulated test |
| INP | How soon the page responds after a tap | Within 200 milliseconds | Only with data from real visitors |
| CLS | Whether the layout stays still or jumps while the page loads | A score under 0.1 | Always, in the simulated test |
The tool always runs a simulated test, and that test gives LCP and CLS for any address. INP is measured on real visitors, so it appears only when Google has data from enough of them, which a new or low-traffic site will not have. In its place the simulated test shows Total Blocking Time: how long the page was too busy to respond while it loaded. If your report has no INP, watch that figure instead.
Test more than the home page. Test a product or service page, and the page your adverts and social links point to, because new visitors often arrive there.
Use a phone the way a customer does. Take a mid-range phone, turn Wi-Fi off, open a private window so that nothing is saved from earlier, and tap the link you share on social media. Do it away from the office, and once more in the evening when the network is busy. Count the seconds until you can read something and tap something.
The usual causes, in the order worth fixing
- Images far larger than the screen. A photo straight from the camera or the designer is shown in a space the width of a phone, and the phone downloads all of it. The fix is to serve a size that fits each screen, in a modern format such as WebP, and to load the images further down the page only when the visitor scrolls to them.
- Arabic web fonts in too many weights. Arabic font files are large, and each weight (light, regular, medium, bold) is a separate file. Two weights instead of six are usually enough for a company site. Choosing the typeface is part of doing Arabic properly, which we covered in True RTL is not a checkbox.
- Third-party scripts. Chat widgets, trackers, sliders and embedded social feeds are each fetched from another company's server and run on the visitor's phone. Remove the ones nobody uses, and load the rest after your own content.
- Page-builder themes that load everything on every page. A theme sold with dozens of features sends the code for all of them to every visitor, whichever page they open. Switch off the modules you do not use. For a site that matters to sales, a theme built for that site is lighter.
- No caching. Without it, the server rebuilds each page for each visitor and the browser is told to keep nothing. With it, the server keeps ready-made copies of pages, and the phone keeps the files that do not change between visits.
- A server far from the visitors. Every request travels to the server and back, and distance adds delay to each one. Ask your host where the site is served from. A content delivery network (CDN) keeps copies of your files in many places and hands each visitor the nearest one, which pays off when your visitors are spread across several countries.
The order matters because the first three are cheap and usually bring the largest gain. A different server or a CDN will not rescue a page that carries a photo of several megabytes and six font files.
How to write speed into a contract
"The site will be fast" cannot be measured, so it cannot be enforced. Ask for a clause that names four things:
- The numbers: LCP within 2.5 seconds and CLS under 0.1 in the simulated test. INP within 200 milliseconds once Google shows real-visitor data for the site, and until then a written target for Total Blocking Time.
- The page types they apply to: the home page, a listing page, a product or service page, and the contact or checkout page.
- How they are measured: on mobile, in PageSpeed Insights.
- When they are measured: at handover, and again after your real content has been entered.
The last point protects both sides. A site is quick with three demonstration photos and slow with three hundred uploaded from a phone. Ask for the content management system to resize images by itself when staff upload them, so that daily use does not undo the work.
For a site you already have, ask for a short speed review instead: measure, list the causes in order of effect, fix the first three, measure again.
Slow pages and lost sales
Slow pages lose visitors. Someone who tapped an advert or a link in a story is one tap away from going back, and you have already paid for the first tap.
Your analytics shows where to begin. Look at which pages people land on and how many of them arrive on a phone. Both are covered in Five numbers your website should tell you every Monday. Start with the heaviest page that receives the most phone visitors, and measure it again after each change so you know which fix earned its cost.
Building and hosting with PearlFibers
Run the two checks on the pages your customers land on. PearlFibers builds fast, bilingual company websites with content management as part of its software development work, and hosts sites on shared plans that can be scaled up or down without moving the site. If the results point at how the site was built, start with the first page. If they point at the hosting, compare the plans on the second.