Why your site feels slow in India even when the speed test says it is fast
You run the speed test, it comes back green, and a customer in Indore still says the site takes a moment to "wake up". Both of you are right. The test and the customer are measuring different things.
The score measures work. The wait measures distance.
A performance score grades how efficiently a page renders once the bytes arrive: image sizes, blocking scripts, layout shift. It says almost nothing about how long the first byte took to turn up. That first wait is TTFB, and on a site hosted far from its audience it is often the largest single number on the page.
The reason is that a browser does not make one trip to your server. Before it sees anything it does a DNS lookup, opens a TCP connection, and negotiates TLS. Each of those is a round trip. A request from Pune to a server in Bengaluru is a round trip of roughly 20-30 milliseconds. To Singapore it is nearer 90. To a data centre in Virginia it is 230 or more.
Multiply that by the handshakes and you have spent close to a second before the server has read a single line of your code. Nothing in your codebase can win that time back.
Why a CDN does not fully solve it
A CDN caches your images, CSS and JavaScript at an edge near the visitor, and that genuinely helps. But the HTML document itself is usually dynamic - a logged-in user, a cart, a price - and dynamic HTML is not cached. So the first and most important request still travels the whole way to your origin server, and the rest of the page cannot start until it comes back.
A CDN in front of a distant origin makes the second half of the page fast and leaves the first half exactly where it was.
What to measure instead of the score
- TTFB from an Indian connection. Not from your laptop on office fibre - use a test tool set to Mumbai, Bengaluru or Delhi. Under 200ms is good. Over 600ms and your visitors are waiting on geography.
- A traceroute from a customer. Ask one to run it. If the path leaves the country and comes back, you have found the problem.
- Largest Contentful Paint on real visits, from Search Console rather than from a lab test. That is the number Google actually uses.
The order to fix things in
- Move the origin closer. Everything else is rounding error next to this. Our shared and VPS plans are served from data centres chosen for latency to Indian visitors.
- Turn on server-side caching so a repeat request is answered from memory instead of rebuilt from the database.
- Then worry about image formats and script loading - which is where the speed score was pointing you all along, and which matters once the first byte arrives quickly.
If you are not sure which of these is costing you, send us the domain. We will run the TTFB test from three Indian cities and tell you what we see, whether or not you move.
