Best Self-Hosted Analytics Tools in 2026
Web and product analytics you run on a machine or VPS you operate, so visitor data stays on your disk.
Umami is the lightest full-featured web analytics stack you can stand up with Docker and a small Postgres database. Matomo is the closest self-hosted stand-in for a Google Analytics-style report set, with the On-Premise edition free of pageview caps. Plausible Community Edition is AGPL and self-hostable; the hosted cloud product is paid. PostHog is the product-analytics pick (events, funnels, flags, replay) with a real self-host path. GoatCounter and Ackee stay tiny. Countly covers web plus mobile from one Community Edition. Pirsch Analytics is mostly a hosted cloud product, included here so that difference is explicit.
Self-hosted analytics means the collector, the database, and the dashboard run on a VPS or machine you operate, not on a vendor's multi-tenant cloud. That is a different product from a "privacy-friendly" SaaS that still stores every hit on someone else's Postgres. Toolradar data: Toolradar lists 10,429 published tools as of September 1, 2026, including 1,232 published analytics tools, 36% of them free or freemium (65 free-only, 376 freemium, 783 paid).
Most of that 1,232 is hosted SaaS: Mixpanel, Amplitude, GA4, heap-style event clouds, and paid cookieless widgets. This guide keeps the eight products you can actually run yourself for site or product analytics. We skipped Fathom because it is a paid cloud product, not a self-host flagship. We also skipped anything that only offers a tracking snippet pointed at a vendor endpoint.
The split that matters is web analytics versus product analytics. Umami, Matomo, Plausible, GoatCounter, Ackee, and Pirsch Analytics count pages, referrers, and (sometimes) custom events. PostHog and Countly treat every click as an event and add funnels, retention, and, in PostHog's case, feature flags and session replay. If you only need "how many people hit /pricing," do not stand up PostHog. If you need "where do trial users drop," Umami will feel thin.
License and database choices are part of the ops bill. Ackee and Countly sit on MongoDB, and MongoDB's SSPL is not an OSI-approved open-source license, even though the apps themselves may ship under MIT or AGPL. Redis, a common cache in heavier stacks, relicensed away from BSD-style OSI terms. Read the LICENSE file for the app and for every datastore you pin in Docker Compose. Self-hosting does not make a source-available dependency "open source."
Top Picks
Based on features, user feedback, and value for money.
| Tool | Starting price | Rating | Best for |
|---|---|---|---|
| Umami | From $9/mo | 4.5(5) | Developers who want a clean, cookieless dashboard and a small ops footprint, not a prod... |
| Matomo | From $19/mo | 4.4(158) | Orgs that want familiar web-analytics reports and a GDPR story they can point at a box... |
| Plausible | From $9/mo | n/a | Teams that like Plausible's small script and want the community edition on their own VPS |
| PostHog | From $750/mo | 4.5(1,054) | Product teams that need event analytics on infrastructure they run, and will pay the op... |
| GoatCounter | Free plan | n/a | Blogs and small sites that want counts without a product-analytics warehouse |
| Countly | From $40/mo | 4.1(20) | Teams that already ship an app and a site and want one self-hosted product-analytics brain |
| Ackee | Free | n/a | Developers who want anonymized, cookieless stats and will operate Node and MongoDB them... |
| Pirsch Analytics | From $6/mo | n/a | Teams comparing cookieless cloud analytics who need an honest note that Pirsch is not a... |
Developers who want a clean, cookieless dashboard and a small ops footprint, not a product-analytics suite
Umami's pricing is incredibly generous, offering a full-featured self-hosted option for free.
Watch out
Cloud overage fees not specified
Orgs that want familiar web-analytics reports and a GDPR story they can point at a box they run
Matomo's pricing is incredibly generous for its On-Premise option, being completely free.
Watch out
Cloud overage fees for exceeding hit limits
Teams that like Plausible's small script and want the community edition on their own VPS
Plausible's pricing is fair, especially for its privacy-focused approach.
Watch out
Overage fees for exceeding pageview limits
Product teams that need event analytics on infrastructure they run, and will pay the ops cost
PostHog's pricing is quite generous, especially with its robust free tier offering 1M events and unlimited seats.
Watch out
Overage fees for events beyond free tier
Blogs and small sites that want counts without a product-analytics warehouse
GoatCounter's pricing is exceptionally generous, offering a fully-featured free tier for non-commercial use.
Teams that already ship an app and a site and want one self-hosted product-analytics brain
Countly's pricing is quite generous, especially with its robust free tiers.
Watch out
Push notifications are add-on
Developers who want anonymized, cookieless stats and will operate Node and MongoDB themselves
Ackee's pricing is exceptionally generous as it is entirely free and open-source.
Teams comparing cookieless cloud analytics who need an honest note that Pirsch is not a self-host flagship
Pirsch Analytics offers a fair starting price with its Starter tier at $6/month for 10K page views, which is competitive for cookie-free analytics.
Other Web Analytics worth considering
Beyond the editorial top picks, these are also strong choices we evaluated.
What Self-Hosted Analytics Actually Means
Self-hosted analytics is software you install on infrastructure you control: a $6 VPS, a homelab box, or a Kubernetes namespace. The tracking script on your site still runs in the visitor's browser, but the beacon hits your origin (or a reverse proxy in front of it), and the rows land in a database you back up. When the disk dies, that is your restore problem. When a regulator asks where EU data lives, you can point at a region you picked.
That is different from a local file. Obsidian on a laptop is local-first notes, not a server you operate. Umami and Grafana are the other pattern: a process you keep alive, patch, and expose over TLS. This guide is the server pattern. None of these eight are "drop a folder on your Mac and you are done," and that is the honest ops cost.
It is also different from "we are GDPR-friendly SaaS." A hosted Plausible or Pirsch Analytics account can be cookieless and still be a processor you have to list. Self-hosting removes that processor for the analytics store, at the cost of you becoming the operator: TLS certs, Postgres upgrades, and a backup that is not "the vendor has replicas."
Why Running Analytics Yourself Still Matters
Visitor and product data is a record of who came, what they clicked, and often which account they were in. Putting that on a US SaaS by default is a legal and a product decision. Self-hosting is how public-sector sites, clinics, and privacy-first indie apps keep a traffic graph without sending every pageview to a third-party ad or product-analytics company.
It also caps the surprise invoice. Hosted Plausible and Matomo Cloud charge on pageviews. Hosted PostHog charges on events and recordings. A busy marketing site can outgrow a cheap cloud tier in a month. A self-hosted Umami or Matomo On-Premise instance has a server bill that stays flat until the disk or CPU actually hurts. The trade is that you, not the vendor, apply security updates.
The third reason is license honesty. Analytics stacks pull in caches and databases that have changed license. Redis moved off a classic BSD-style OSI license. MongoDB uses SSPL, which OSI has not approved. Sentry (covered in the monitoring guide) left OSI terms for FSL and BSL-style source-available licenses. If "we only run open source" is a procurement rule, you have to read past the marketing word "open" on the homepage.
Key Features to Look For
The script or SDK must send hits to a URL you control. If the only supported endpoint is the vendor's cloud, the product is not self-hosted analytics.
Official Compose files or a single binary with Postgres or SQLite beat a wiki page that assumes you already know the stack.
AGPL, GPL, and MIT on the app are OSI-approved. SSPL (MongoDB) and relicensed Redis are not. FSL and BSL-style licenses are source-available, not OSI open source.
Pageviews and referrers are one job. Event funnels, feature flags, and replay are another. Pick the job before you pick the container.
Self-hosting does not auto-solve consent. You still decide whether the script sets a cookie and what you write in the privacy notice.
SQL dump, CSV, or an API you can cron. Hosted tiers expire history; your disk does not, unless you set a retention job.
Mistakes to Avoid
- ×
Calling a cookieless SaaS 'self-hosted' because the script is small. If the beacon hits the vendor, you are not running analytics yourself.
- ×
Standing up PostHog for a brochure site. Event warehouses are the wrong shape for 'top pages this week.'
- ×
Ignoring datastore licenses. Ackee and Countly on MongoDB inherit SSPL on the database. Redis relicensing bites stacks that pin an old 'BSD Redis' assumption.
- ×
Skipping backups because Docker 'has volumes.' The volume is the product. Test a restore.
- ×
Putting Fathom in a self-hosted shortlist. It is paid cloud, not a self-host flagship, which is why it is not in this list.
Expert Tips
- →
Start with Umami or GoatCounter on a throwaway VPS, record one real site for a week, then decide if you need Matomo reports or PostHog events.
- →
If legal asked for 'no US processor,' write the data-flow diagram first: script, your origin, your Postgres or MongoDB, your backups. Then pick the tool that matches that diagram.
- →
Keep analytics off the same box as production without a snapshot plan. A full disk on the shop database and the hit store at the same time is a bad night.
- →
Re-read licenses once a year. Redis relicensed. MongoDB is SSPL. App licenses move too. The homepage word 'open' is not the LICENSE file.
- →
For product analytics, budget RAM for PostHog or Countly the way you would for any app, not the way you would for a 128 MB Umami toy install.
The Bottom Line
If you want web analytics on a VPS you operate, start with Umami or Matomo On-Premise. Use Plausible Community Edition when you want that UI and will run the AGPL stack yourself; the cloud product is paid and separate. Use PostHog or Countly when the job is product analytics, not pageviews. GoatCounter and Ackee stay small. Pirsch Analytics is mostly cloud. From 10,429 published tools, these eight are the self-hostable analytics set we would actually install in 2026.
Frequently Asked Questions
What is the best self-hosted analytics tool in 2026?
Is Plausible self-hosted or only paid cloud?
Both exist, and they are different products. Plausible Cloud is paid SaaS. Community Edition is AGPL and you run it on your own machine or VPS. Signing up on the marketing site is not self-hosting.
Why is Fathom not on this list?
Fathom is a paid cloud analytics product. It is not a self-host flagship, so it does not belong in a guide about collectors and databases you operate.
Does self-hosting analytics make MongoDB or Redis 'open source'?
Is Pirsch Analytics self-hosted?
Pirsch Analytics is mostly a hosted cloud product. Treat it as a privacy-oriented SaaS comparison, not as the Docker compose you would pick instead of Umami or Matomo.
Related Guides
From the team behind Toolradar
Editorial content for B2B tech
We measure pipeline contribution by channel, not just clicks and sessions.
See how we workReady to Choose?
Compare features, read reviews, and find the right tool.