Best open source analytics tools. An honest roundup.
Six open-source analytics tools worth running, ranked by the job they do best, with each one's real tradeoff stated plainly. No trashing, no invented flaws, and a straight answer on which to pick.
How to choose in one line
- ·Want to ask your numbers in plain english and trust the answer? smolanalytics: one Go binary, a verdict on what to fix, answers computed not guessed.
- ·Want the deepest all-in-one product suite? PostHog: funnels, replay, flags, and experiments together, if you can carry the self-host weight.
- ·Just need clean, private web analytics for a site? Plausible or Umami: cookieless, tiny, no consent banner.
- ·Want a mature, do-everything web suite? Matomo, if you are fine running PHP and MySQL.
- ·Mobile-heavy or enterprise? Countly, for its SDKs and community edition.
The tools, ranked by what they are best for
Ranked for the specific reader this page serves: someone who wants a lightweight, honest, ownable analytics tool they can ask questions of. If your priority is a different job (say, session replay), the order would shift, and the tradeoffs below say exactly when it should.
- 1smolanalyticsask-in-plain-english · one binary · verdictMIT
Best for: Best if you want to ask your real numbers in plain English and get a computed answer, not build dashboards. Web and product analytics (visitors, referrers, funnels, retention, paths, cohorts) from one snippet, plus a daily verdict on what to fix.
Tradeoff: It deliberately does not do session replay, feature flags, experiments, heatmaps, or surveys. If you need those, keep a heavier tool for them. It is a young, one-person project.
- 2PostHogall-in-one product analyticsMIT (with some paid features)
Best for: Best all-in-one open-source product suite: funnels, session replay, feature flags, A/B experiments, surveys, and heatmaps in one platform. If you want every product tool under one roof, it is the deepest.
Tradeoff: Correspondingly heavy to self-host. Running it yourself means Kafka, ClickHouse, Redis, and Postgres; PostHog itself has written that many teams lack the resources to run it reliably.
- 3Plausible (self-host)privacy-first web analyticsAGPL
Best for: Best lightweight, privacy-first web analytics. Cookieless by default, no consent banner, a clean single-page dashboard, and a script that is a fraction of a kilobyte. Easy to self-host with its Docker setup.
Tradeoff: Web analytics only: it is not built for product events, funnels, retention, or cohorts. Self-hosting runs on Elixir with ClickHouse and Postgres behind it.
- 4Umamisimple, self-hosted web analyticsMIT
Best for: Best minimal self-hosted web analytics. Privacy-friendly, cookieless, a tidy dashboard, and a very small footprint. A popular free swap for Google Analytics on a marketing site.
Tradeoff: Also web-first: event tracking is basic, and there are no real product funnels, retention, or cohorts. It needs a Postgres or MySQL database alongside it.
- 5Matomomature full web-analytics suiteGPL
Best for: Best mature, full-featured web-analytics suite. The long-standing open-source alternative to universal Google Analytics, with the deepest reporting, goals, ecommerce, and a big plugin marketplace.
Tradeoff: Heaviest of the web-analytics tools to run: a PHP + MySQL application to install, host, and keep updated. The interface is dense, and it is web-analytics-shaped, not product-analytics-shaped.
- 6Countlymobile + enterprise product analyticsAGPL (community) / commercial
Best for: Best for mobile-heavy and enterprise product analytics. Strong SDKs across mobile platforms, crash reporting, push, and a community edition you can self-host.
Tradeoff: The community edition is a subset of the enterprise product, and the stack (Node + MongoDB) is more to operate than a single binary. Aimed at bigger teams than a solo builder.
How they compare at a glance
| tool | shape | self-host footprint | ask in plain english | license |
|---|---|---|---|---|
| smolanalytics | web + product | one Go binary, no external DB | yes, computed not guessed | MIT |
| PostHog | product (all-in-one) | Kafka + ClickHouse + Redis + PG | via generated SQL | MIT-ish |
| Plausible | web | Elixir + ClickHouse + PG | no | AGPL |
| Umami | web | app + Postgres/MySQL | no | MIT |
| Matomo | web (full suite) | PHP + MySQL | no | GPL |
| Countly | product (mobile-first) | Node + MongoDB | no | AGPL / commercial |
Footprints are the typical documented self-host stacks; each project's cloud hides all of it for you. "Ask in plain english" means a first-class natural-language surface, not that you can query a database yourself.
Where smolanalytics fits, and where it does not
smolanalytics is a single MIT-licensed Go binary, stdlib only, roughly 7 bytes per event, with no Kafka, ClickHouse, or Postgres to run beside it. From one snippet you get web analytics (visitors, referrers, UTM) and product analytics (funnels, retention, paths, cohorts), plus a verdict that tells you what to fix on the dashboard and in a morning brief.
The wedge is the ask surface. You ask in plain English from a dashboard bar, or from your own Cursor or Claude over MCP (47 tools, 13 prompts), and the answer is computed from the same deterministic reports the dashboard renders, never generated by an LLM. A CI agreement test fails the build if the editor's answer ever differs from the dashboard's, so a number can't be hallucinated. You bring your own AI model, so the AI part is free. It has a cookieless mode with no consent banner, and importers from PostHog, Umami, CSV, and JSONL so switching is a copy, not a rebuild.
The honest limit: it deliberately does not do session replay, feature flags, experiments, heatmaps, or surveys. If those are central to how you work, PostHog is the better fit and this page says so plainly. It is also a young project built by one person. Pick smolanalytics when what you want is a straight, ownable, trustworthy answer on what to fix, cheaply.
docker run -p 8080:8080 ghcr.io/arjun0606/smolanalytics