Performance advice is mostly written for users on fast connections close to the server. Neither assumption holds for a typical Egyptian visitor, which is why sites that score well in testing tools still feel slow in Cairo. Three things decide real-world speed, and they are not the three that lab tools emphasise.
1. Distance beats optimisation
If your server sits in Europe or North America, every request pays a round trip that no amount of code optimisation removes. The connection setup alone — DNS, TCP, TLS — costs several round trips before a single byte of content moves. A CDN with presence near your users fixes this for static assets, and it is the single highest-leverage change available: it addresses the part of the delay that optimisation cannot touch.
What a CDN does not fix is the dynamic response. If every page is generated per request against a distant database, the CDN serves your images quickly and your HTML slowly. Static generation or edge caching of the HTML itself is what closes that gap.
2. Mobile networks are not slow, they are variable
The common assumption is low bandwidth. The real characteristic of mobile networks is variance: latency swings, packets drop, and a connection that measured well a second ago stalls. This punishes designs that need many sequential requests before anything renders — each round trip is another chance to hit a bad moment. Fewer, larger requests survive variance better than many small ones, which inverts some older optimisation advice.
It also means averages mislead. A page that loads in two seconds on average may take twelve at the ninetieth percentile, and that tail is a real share of your visitors, not an outlier to be dismissed.
3. Page weight, and where it actually goes
- Images are almost always the largest cost, and almost always the easiest win. Serving a 2,000-pixel image into a 400-pixel slot is the most common single waste on the web.
- Fonts are the quietest cost. An Arabic-English site can load two families across several weights without anyone noticing, and text stays invisible while they arrive.
- Third-party scripts — analytics, chat widgets, tag managers, ad pixels — are the cost nobody owns. Each is added by a different person for a defensible reason, and collectively they often exceed the site itself.
How to measure honestly
Lab tools run on a fast machine on a fast connection and tell you what is theoretically possible. Field data tells you what your users experience. Prefer the second. Test on a mid-range Android phone on a mobile connection, not on the office WiFi on a laptop — and test the pages people actually land on from search and ads, not just the homepage, which is usually the most optimised page on any site.
Abdulrahman Sayed
EdTech & HealthTech Insights