Why your website is slow on a mid-range Android phone
Most performance testing is done on a fast laptop over office wifi. Your customers are on a ₹12,000 phone on patchy 4G, and that gap is where conversions quietly disappear.
Every site we are asked to fix has the same origin story. It tested well, it launched, and then enquiries were lower than expected with no obvious reason. The reason is almost always that nobody tested it the way customers actually use it.
The testing gap
A development laptop has a fast processor, a wired connection and a warm cache. A customer in Nagar Kurnool has a mid-range Android phone, an intermittent 4G signal and no cache at all because it is their first visit. The same page can take under a second on one and eleven seconds on the other.
Eleven seconds is not a slow page. It is an abandoned one.
What to measure instead
Test on a real mid-range device, throttled to slow 4G, with the cache cleared. If you only change one thing about how you evaluate your site, make it this.
The four things that usually cause it
In our experience the same four problems account for most of the damage.
Images sized for a desktop
A 2400-pixel-wide hero image served to a 390-pixel-wide screen wastes most of the bytes it costs. Serving modern formats at appropriate widths routinely cuts page weight by half.
JavaScript that blocks the first paint
Analytics, chat widgets, tag managers and font loaders all compete to run before anything appears. Each one is defensible on its own; together they push first paint past the point where people leave.
Fonts that hide the text
A custom font loaded without a fallback strategy leaves the page blank while it downloads. The text was ready in 200 milliseconds and the visitor saw nothing for three seconds.
Server response time
Shared hosting under load, an unindexed query on the home page, or an API call made synchronously during page render. These are invisible on an idle staging server and obvious the moment real traffic arrives.
Where to start
Fix images first. It is the least invasive change and usually the largest single gain. Then defer non-essential scripts, then look at the server. Re-test on the real device after each change rather than at the end, so you know which change actually helped.
Laxman
Founder & Principal Engineer
Founder of Kaarvex Technologies. Builds data platforms and business systems, and writes about the parts of software delivery that are usually left out of the case study.