How a PWA works
A progressive web app is still a website. What makes it behave like an installed app is two extra files and a security requirement.
The first is a manifest: a small description of the app — its name, its icon, the colours, whether it opens with or without browser chrome — which lets a phone add it to the home screen. The second is a service worker, a script the browser keeps running in the background even when the site is closed. It sits between the page and the network, so it can serve pages it has already stored when the connection is poor or absent, and it is what makes push notifications possible. Both require the site to be served over HTTPS.
Support is not even. Android handles installation and notifications comfortably; on iPhone the behaviour is more limited and depends on the visitor adding the site to their home screen first.
Why a PWA matters
It removes the app store from the equation. There is no review process, no separate build for each platform, no download before a first visit, and no second codebase to maintain — you ship an update to your website and everyone has it. For a business that wants repeat use rather than a genuinely native experience, that is a large saving.
Offline capability is the part that earns its keep in practice. Where connections drop — a trekking region, a hospital basement, a bus in Kathmandu traffic — a stored page that opens instantly is worth more than a polished one that will not load. Repeat visits also feel quicker, because the shell of the site is already on the device.
Where a PWA goes wrong
The biggest risk is search visibility, and it comes from how PWAs are usually built rather than from the idea itself. Many are single-page applications that render everything with JavaScript, which is a well-known source of indexing problems: content that never appears in the crawled HTML, navigation that is not made of real links, and pages that have no distinct URL to rank.
Stale caching is the second. A service worker that holds pages too eagerly will keep showing an old price, an old opening time or a withdrawn offer long after you corrected it, and it will do so only for the visitors who came back — the ones you least want to mislead.
The third is asking for too much too early. A prompt to install and a request to send notifications, thrown at a first-time visitor before they have read anything, is refused far more often than accepted, and the refusal is difficult to reverse.
How to act on it
Start from a real need. Offline access, quick repeat visits and notifications are the reasons to build one; a slow first load is not, because a service worker does nothing on the visit that matters most.
Keep every page a real, server-rendered, indexable URL, and treat the app behaviour as a layer on top rather than as the site itself. Give the service worker a clear update rule so a fresh version arrives promptly. Delay any install or notification prompt until someone has shown genuine interest. And before committing, check the ordinary mobile experience first — most of the speed complaint a PWA is meant to solve is usually cheaper to fix as ordinary web development work.
