Self-hosting is an ops decision
Every tool here is open source and self-hostable, so the question isn’t whether you can own your data; it’s what running the tool costs you in operations. Weigh three things:
- Footprint: how many services you have to stand up and keep alive.
- Infrastructure: a lightweight script + one database, or a columnar event store and a queue.
- Maintenance: updates, backups, and scaling as your volume grows.
This roundup focuses on the self-hosting experience. For a feature-by-feature look at what each tool measures, see the best open source analytics tools.
Web analytics, lightest to run
Umami
A simple, cookieless web-analytics tool on Node with a single Postgres or MySQL database, among the easiest to run, comfortably on a small VPS. Umami v2 added funnels, retention, and journeys, but it stays aggregate (no per-person profiles). MIT-licensed. Best for: a minimal, low-maintenance traffic dashboard. If you outgrow aggregate stats, see Umami vs product analytics.
Plausible
Privacy-first, cookieless web analytics with a famously light script. The Community Edition self-hosts on Elixir with Postgres and ClickHouse, a bit more to run than Umami, still modest. AGPL-3.0. Best for: clean traffic analytics without cookie banners. Compare it as a Plausible alternative.
Matomo
The most full-featured open-source web analytics platform and the closest direct Google Analytics replacement, on PHP + MySQL. More capability means more to operate, and heatmaps and recordings are paid plugins. Best for: a complete GA replacement for web and marketing. See the Matomo alternative.
Product analytics, heavier, but behavioral
PostHog
Open-source (MIT) and very broad, analytics, replay, flags, experiments. The catch for self-hosters: the open-source build is a community “hobby” Docker Compose deployment (one project per instance, no commercial support), and PostHog steers production users to its cloud. Operationally it’s the heaviest option here. Best for: teams that want breadth and will accept the ops. Details: PostHog alternative.
Pug
Open-source (AGPL-3.0) product analytics with unified profiles and autocapture. It’s built to be self-hosted: the API and every worker run as a single Go binary, backed by Postgres, ClickHouse, and NATS, fewer moving parts than a multi-service platform, so the operational surface stays small. Self-hosting is free forever, with no event caps, and tracking defaults to cookieless. Nothing is written to the device until you grant consent. It’s new, out of beta since October 2026. Best for: product analytics you own without service sprawl. See self-hosting Pug and the setup guide.
How to choose
If you only need website traffic, the lightest tools (Umami, Plausible) are the easiest to run; Matomo is the fullest GA replacement if you accept more ops. If you need product behavior: funnels, retention, and a profile per person: you’re into product analytics, where the event store makes the stack heavier; Pug keeps that footprint smaller by shipping as one binary. Whatever you choose, self-hosting keeps every event on your own infrastructure. Moving off Google specifically? Our free Google Analytics replacement guide walks through it.