How State of Indie is measured

Every figure in the report is read from the outside of real products — the way a visitor, a search engine or a mail server meets them — and frozen once a month. This page says exactly who is counted, where each number comes from, and what each source cannot see.

Who is counted

The population is every product submitted to Awesome Indie that passed moderation: launched products, and accepted products waiting for their launch day. Products by banned accounts are left out. In the November 2026 edition that is 8,560 products.

Website figures count each web address once. A product whose page lives on someone else’s platform — an app store, GitHub, a marketplace — has no website of its own to measure and is left out of them, and so is a product that is one page deep inside a bigger site (that site’s rank, links and mail are not its own), a site that did not answer its last check, and one that turned out to be a parked domain. When several products share an address, it is counted through the earliest one. That leaves 7,273 websites this edition, 3,542 of them with real-user speed data.

Product figures (pricing, platforms, categories, social links) count every product in the population that states the field — filled in by the maker, or read from the product’s own website when the maker left it blank. An empty field means “not stated” and is never guessed. Because those fields are read from the site, the products that state them are mostly products whose site works, and recent products state them more often; every chapter says how many products its figures rest on.

Survival is the one chapter that also counts launches taken down since their launch day — leaving them out would flatter the result, because a product is often removed because its site died. That chapter covers 6,605 launches.

This is not a random sample of the web or of startups. It is the products makers chose to launch on a directory for indie software, which leans towards small, recent, often solo-built products — and towards makers who want to be found.

Where the numbers come from

Site check (our own crawler)

One plain request to every product’s web address each month: status, redirects, response headers, TLS certificate, HTTP version, robots.txt, sitemap.xml, llms.txt, and the technologies visible in the HTML, scripts, headers and cookies.

Limits: Sees what a visitor’s browser is sent, not what runs on the server. A site behind a bot wall answers with a challenge page and is counted as checked but not detected.

Used in: Frameworks, Website builders, CSS, Hosting, Analytics, Website health, AI crawlers

DNS

MX, SPF and DMARC records of every domain the product owns.

Limits: Products on a free subdomain (vercel.app, github.io…) do not own their domain’s mail and are left out of the email chapters.

Used in: Email providers, Email security

Chrome UX Report

Google’s public real-user data: the 75th percentile of LCP, INP and CLS over 28 days, all devices, per origin; and the monthly CrUX top-list popularity buckets.

Limits: Only origins with enough Chrome traffic appear, so the speed figures describe the busier half of indie sites. Published under CC BY 4.0 by Google.

Used in: Core Web Vitals, Hosting speed, Popularity, Does speed matter?

PageSpeed Insights (Lighthouse)

One simulated mobile load of the home page: Performance, Accessibility, Best Practices and SEO scores.

Limits: A lab test, not visitors; one run per month, so a single slow moment can move one site’s score.

Used in: Core Web Vitals, Hosting speed, Does speed matter?

Tranco

A research ranking of the top million registrable domains, combining several traffic lists.

Limits: Matched on the exact domain only, so products on subdomains or shared hosts are never ranked.

Used in: Popularity, Does speed matter?

Common Crawl web graph

The domain-level link graph of Common Crawl’s open crawl: how many distinct domains link to each product’s domain, and its harmonic-centrality rank.

Limits: A sample of the web, so counts are far below what commercial link indexes report — they are for comparing products with each other, not absolute totals. Refreshed with each Common Crawl release.

Used in: Backlinks, Does speed matter?

Internet Archive (Wayback Machine)

The date the product’s address was first captured.

Limits: A domain bought second-hand carries its previous owner’s history, so the oldest dates say more about the domain than the product.

Used in: Time to launch, Popularity, Backlinks

Hacker News, Reddit, Bluesky and Lobsters

Public posts that link to the product’s domain, with their points and comments.

Limits: Matched by link, so a product posted without its address is missed; products living on a page of a bigger site are left out of that chapter.

Used in: Mentions

GitHub

Stars, forks, licence and last commit of the repository a product links to.

Limits: Only products that link a repository.

Used in: Open source

What makers state

Pricing model, starting price and period, free trial, platforms, categories, social links and the maker’s country, as written on the product page.

Limits: Only what is stated is counted; an empty field is “not stated”, never guessed. Prices are compared in US dollars only.

Used in: Pricing, Platforms, Categories, The AI wave, Countries, Social media

Similarity model

Text embeddings of every launched product’s description, with a model judging which close neighbours actually do the same job.

Limits: Covers launched products only; a niche is a group of products judged interchangeable, labelled from the words they share.

Used in: Crowded niches

The rules every chapter follows

  • No share or median is printed for a group of fewer than 30. Smaller groups are dropped from charts and shown greyed in tables, so a reader sees they exist without taking them as a finding. Trend lines start only where every period has at least 60.
  • Every share says what it is a share of, and every chapter prints its n.
  • Technologies count only when the detection is confirmed — a header, a script or a fingerprint that cannot belong to anything else, or a framework that necessarily implies another (Next.js implies React). A mention in the page text alone does not count.
  • Trends “by when the product was submitted” describe products as they are today, grouped by the half-year they were submitted; they show how newer products differ from older ones, not how any one product changed.
  • The report never uses anything that happens on Awesome Indie itself — upvotes, views, comments or launch order — so a product cannot move a figure by being popular here.
  • No figure names a product as slow, insecure or failing. Individual products appear only as examples of a group.

Editions, corrections and history

On the first of every month, after the night’s checks, every chapter is recomputed from one fresh read of the population and stored as that month’s edition. It is never recomputed afterwards: the chapter URL always shows the newest edition, and every older one stays readable from the edition menu (?edition=YYYY-MM).

A headline figure that moves by more than a few points from one month to the next is checked by hand before anyone reads anything into it, because it is more often a detection rule that improved than a change in the world. When a rule changes, the chapter’s “How this was measured” section says so, and the edition-by-edition chart shows the step.

Found something wrong? Write to support and it will be corrected in the next edition.

Licence and citation

The figures are published under CC BY 4.0. Quote them, chart them and republish them, with credit to State of Indie by Awesome Indie and a link to the chapter. Chrome UX Report data is Google’s, also under CC BY 4.0; Tranco and Common Crawl are credited where used.

Every chapter has a ready-made citation, a CSV of its main table and embed codes for its charts. All the files are on the data page.