Analytics cookies only (Google Analytics, PostHog). No ad networks, no cross-site tracking. See our privacy policy.
upstream: github.com/arp242/goatcounter (opens in new tab) · github.com/umami-software/umami (opens in new tab)
Pick Umami for the easiest cookieless Google Analytics swap that still feels like analytics — a clean single-page dashboard with custom events that carry properties, funnels and retention, MIT-licensed, two containers (Next.js + PostgreSQL) on 1 GB. Pick GoatCounter when running as little as possible is the requirement: one Go binary or container writing to a SQLite file, no database server, named events but no properties, a JavaScript-free tracking pixel and log-file import, and a README that says it does not need a GDPR notice — at the cost of a narrower feature set and the less familiar EUPL-1.2 license.
GoatCounter and Umami are the two featherweights in this hub — the picks for someone who wants to know how the site is doing and wants nothing else to run. Both are cookieless, both fit in one short compose file, and both sit far below the 2 GB floor of the ClickHouse-backed tools. They differ on how small "small" is, and on how much dashboard you get for it.
GoatCounter is a single statically compiled Go binary — or the project's own Docker image — writing to a SQLite file in one named volume. There is no second container. The README says the release binaries "contain everything you need" and that "the only dependency is somewhere to store a SQLite database file or a PostgreSQL connection." It publishes no RAM figure; the 256 MB on our page is our estimate, and it is generous.
Umami is a Next.js application that wants a PostgreSQL container beside it. Two containers, one compose file, 1 GB on our page. Still tiny by the standards of this hub, but it is an application with a database, where GoatCounter is a process with a file.
Umami gives you the modern single-page analytics dashboard: pageviews,
unique visitors, referrers, top pages, countries, devices — plus custom
events that carry data properties, and the funnels, retention and heatmaps
its MIT build includes. That is the layer that separates "a counter" from
"analytics" for most small projects. One script tag on the site and you are
done; a backend can also POST events to /api/send.
GoatCounter is candid about its scope. Its README positions it for people
"confused by the myriad of options and flexibility of Google Analytics and
Matomo that you don't need", and what it tracks is: unique visits (via a
non-identifying hash, no cookies), referring sites and campaigns, browser,
location and screen size, and named events — a count() call with
event: true, a path and a title, but no arbitrary properties. What it adds
that Umami does not is log-file import from nginx, Apache, Caddy or
CloudFront; its JavaScript-free tracking pixel counts pages on your own
site, while Umami v3's pixels are separate stats pages meant for emails and
third-party sites. Both also accept events from your backend. For a documentation site, a blog, or
anything whose readers run script blockers, that flexibility is the point.
Both avoid cookies. GoatCounter goes further in what it claims: the README says it "doesn't need a GDPR notice", and the project's GDPR page explains why — "no personally identifiable information is collected" — while adding the word probably and a reminder that the author is not a lawyer. Umami is cookieless and privacy-focused, and leaves the legal conclusion to you.
Umami is MIT. GoatCounter is EUPL-1.2 — the European Union Public Licence, with a modified "Compatible Licenses" appendix that the LICENSE file says is otherwise identical to the standard text. Both are free-software licenses; MIT is the one nobody has to look up.
Umami wins this comparison for most self-hosters because events with properties, funnels, retention and a dashboard people actually read are worth one extra container, and 1 GB is the smallest VPS tier anyway. GoatCounter is the better pick when minimalism is the requirement rather than a preference: it is the only tool in this hub that runs as a single process on a SQLite file, counts visitors with no JavaScript at all, and says in plain words what it does not do.
GoatCounter lists a lower minimum: 256 MB vs 1024 MB for Umami.
No. GoatCounter is EUPL-1.2, Umami is MIT.
Pick Umami if you want events with properties, funnels, retention and a modern dashboard your teammates will read without a tutorial, and you are fine running PostgreSQL beside it — it is MIT-licensed and runs on 1 GB. Pick GoatCounter if you want the smallest possible thing: a single Go binary or container on a SQLite file, named events, a JavaScript-free tracking pixel, and log-file import, with a narrower feature set by design.
Its README says it "doesn't need a GDPR notice", and the project's GDPR page explains the reasoning — no personally identifiable information is collected, and unique visits are counted with a non-identifying hash rather than a cookie — while adding the word probably and noting the author is not a lawyer. Umami is cookieless too; it leaves the legal conclusion to you.
GoatCounter. It is one statically compiled binary (or one container) writing to a SQLite file, with PostgreSQL optional for large sites; the project publishes no RAM minimum, and the 256 MB on our page is our estimate. Umami is a Next.js app plus a PostgreSQL container; it publishes no figure either, and the 1 GB on our page is ours. Both fit on the smallest VPS tier.
Paid link — we earn a commission if you shop through it.
The full GA replacement vs. the privacy-first dashboard.
Funnels, journeys and replays on both — two containers on 1 GB, or six on ClickHouse and 2 GB.
Two AGPL newcomers on ClickHouse — one replacing Google Analytics, one replacing Mixpanel.