Method. The index is never finished: an always-on engine keeps discovering candidates from
funding and startup news (Inc42, Entrackr, Tracxn, YourStory, ET, Mint, Moneycontrol, TechCrunch India, The Ken,
VCCircle — via RSS and Google News), Wikipedia's India company categories (websites from Wikidata), every NSE
listing since 2005, the public Y Combinator API, a curated seed of new-age Indian companies, and a continuous
lap over the Tranco top-1M traffic list. Company names from news are resolved to websites through free lookups. Each candidate's
homepage and first-party JavaScript bundles are fetched live and matched against 92 vendor
signatures. Detected vendors are priced from published rate cards, scaled by the company's
traffic tier and weighted by how reliably each signature implies a paid contract. Per-seat
engineering tooling is excluded from the headline figure: it needs headcount, which is known
for only a small share of these companies, so including it would rank those rows on a term
the rest cannot have.
Flags. unverified marks a company found by the .in domain sweep and
classified automatically rather than drawn from a known-company source — the classifier
removes foreign brands' India sites, government portals, legacy corporate groups and spam, but
these rows are worth a look before you act on them. infra only means no separately-billed
vendor was detected and the estimate rests on the traffic-tier infrastructure baseline alone;
only companies large enough for hosting to clear $10k on its own are admitted on that basis.
Limits. These are modelled estimates from client-side signals. Server-side spend
(data warehouse, internal APM, ERP) is invisible to a page fetch, and mobile-app-first companies
are under-measured — so the model under-counts more often than it over-counts. Treat this
as a ranked prospect list with evidence attached.