What Pune business websites are built on

What we read, and what we did not

An earlier survey of 9,779 Pune businesses recorded which of them had a website: 989. That survey did not fetch the sites again. So this is a new measurement over the same list. After tidying the addresses there were 853 distinct hosts, and on 2026-10-04 we asked each one for its homepage and read what came back.

Two dates, 87 days apart. The businesses and their websites were recorded on 2026-07-09. The homepages were fetched on 2026-10-04. Every figure describes a site as it was on the fetch date, for the hosts that were on record in July.

The method in short: one request per site to the homepage, secure address first, then the plain one, following a handful of redirects and reading only the start of the page. We looked for the markers a platform leaves in its own pages. We did not log in, crawl inner pages, or test anything beyond that first answer. The request said who we were and gave a contact page.

The first finding is how many sites do not answer

Before any question about platforms, there is a plainer one: did the site come to the door? Of 853 hosts, 647 answered with a usable page, which is 75.8%. The remaining 24.2% did not, and they fail in different ways.

• Unreachable (104): the name or the connection failed outright. This is where the lapsed domains in our DNS research show up.
• Timeout (47): nothing came back inside the time we allowed. A slow server and a dead one look the same from here.
• HTTP error (49): the server answered with a refusal or a failure. The kinds are in the next table.
• Parked (6): the page matched a parking or for-sale pattern. The business has a domain and no site.

Read the error table with care. 27 of the 49 were a 403, a refusal, and some sites refuse any visitor they do not recognise. So the HTTP error count is a floor on broken sites, not a count of dead ones. The 404s and server errors are the clearer cases of a site that is genuinely failing.

For a business, this is the whole point of the section. About one site in four in this list did not give a usable homepage to a polite visitor on the day we asked. Nobody tells the owner. The site is simply not there when somebody tries it.

What the ones that answered are built on

Platform was read for the 647 hosts that answered, and every share below is a share of those 647, not of all hosts attempted and not of all businesses.

WordPress is the clear leader among the sites we could name: 190 of 647, or 29.4%. Next.js, the framework this site is built with, accounts for 35 (5.4%), and other JavaScript frameworks for 24 more (3.7%). The hosted site builders and shop platforms we looked for (GoDaddy, Shopify, Wix, Webflow, Squarespace and Google Sites) together hold 46 sites, 7.1% of those that answered.

4 of the 647 are not a site of the business at all but a profile on a social or directory service. Take those out and the denominator becomes 643; WordPress then reads 29.5%. The difference is small, and we show it so the base is never in doubt.

The largest row is not a platform. 51.9% of the sites that answered showed no marker we recognised. That means we found nothing, not that somebody built the site from scratch. A site can strip its markers, sit behind a service that rewrites its pages, or be plain files with nothing to find. Read that row as "unknown", and read every other row as a floor, because a site we could not identify might be on any of them.

Because of that, the honest summary is narrower than a headline. Among the sites where something could be named, WordPress dominates. Among all the sites that answered, the single biggest group is the ones we could not name.

The earlier survey in this series also reported platforms, from a different probe that named no JavaScript frameworks. Compared only across the five platforms both probes recognise, the two agree: WordPress was 83.8% of them then (227 of 271) and is 81.5% now (190 of 233). Its share of every site we could name is lower here only because this probe also names the framework-built sites the earlier one could not.

Secure and mobile-ready, mostly

• An https request worked on 603 of 647 sites that answered (93.2%).
• Where https worked and the plain address also answered, the plain address redirected to https on 502 of 588 (85.4%). The rest leave a visitor who types the older address on the unsecured version.
• Of the 44 sites where https did not work, 14 (31.8%) failed in a way that looked like a certificate problem. The others could not be told apart from having no https at all.
• A mobile viewport tag was present on 614 of 647 (94.9%).

That is a better picture than the failure rate suggests: the sites that exist and answer are mostly set up for phones and mostly served securely. The gaps are in the minority that are not, and in the sites that did not answer at all. The viewport tag is a proxy. It shows that someone told the browser how to scale the page, not that the page is pleasant on a small screen.

What this does not say

It does not say which platform is better. WordPress leading the named group is a fact about what these businesses chose, or were sold, and says nothing about whether it was the right choice for any of them. For a plain account of when its underlying language is a good fit, see our guide to PHP.

It also does not say anything about how a site performs, how it ranks, or whether anyone visits it. Those are different measurements, and some of them we have made on other pages.

What this does not cover

• The business list and the fetch are 87 days apart. A site that changed, moved or lapsed in between is counted in the July list and measured in its October state.
• Platform detection reads markers in the homepage. It undercounts sites that strip them, sit behind a rewriting service or run headless, so unidentified means no marker found, not custom built. Every named platform share is a floor.
• A page showing markers of two platforms is counted once, under whichever is checked first.
• Only the homepage is read, and only the start of it. A marker further down, or only on inner pages, is missed.
• This is one fetch from one network on one day. A transient failure is counted as a timeout or unreachable and was not retried.
• Sites that block unfamiliar visitors appear as errors, and sites that serve a different page to automated visitors appear as whatever that page is. The error count is a floor on broken sites, not a count of dead ones.
• Parking detection works by phrase and by provider and will miss unfamiliar parking pages, which are counted as answered.
• The viewport tag is a proxy for mobile readiness, not a test of it.
• These are businesses recorded in OpenStreetMap in one district of one city. Coverage varies by area and sector, and a business that is not mapped is not here.

Sources

• Pune business survey, 853 distinct hosts derived from 989 recorded websites. One homepage request per host, secure address first, with a limited number of redirects and a limit on how much of the page is read. Platform is read from the markers a page carries, first match wins. Every percentage names its denominator. Business records from OpenStreetMap, licensed ODbL. Run: scripts/pune-stack.mjs, 2026-10-04 15:27 UTC.
• OpenStreetMap copyright and licence, OpenStreetMap Foundation. Retrieved 23 August 2026.

Revisions

• 4 October 2026 First published.

This page is revised in place rather than replaced, so its address does not change.