Most lists of Google Analytics alternatives are lists of other hosted services. You move your visitors' data from Google's servers to somebody else's and call it privacy. This is the other list: the open-source platforms you run yourself, where the data stays on your infrastructure because it never leaves it.
Web analytics is unusually well served by open source right now, and the three tools below are not hobby projects. It is also a category where the honest comparison includes one thing Google Analytics does that none of them replace, covered at the end.
Why people leave
Three reasons, and they arrive in a different order than they used to.
GA4 itself. The forced migration from Universal Analytics left a lot of people with a tool they never chose, an event model they find opaque, and reports that answer questions they were not asking. A significant share of "Google Analytics alternative" searches are not about privacy at all. They are about wanting the old simplicity back.
The consent banner. Analytics that sets cookies and ships data to an ad-tech company is exactly what consent regulation was written for. The banner costs you a measurable share of your data (everyone who clicks no) and a less measurable share of goodwill. Tools designed to work without cookies change that calculus.
Where the data lives. European data protection authorities have at various points found Google Analytics transfers unlawful, and while the EU-US Data Privacy Framework changed the legal landscape in 2023, plenty of teams concluded they would rather not have their traffic data be a compliance question at all. Self-hosting removes the transfer, because there is none.
The three worth knowing
All three are open source, actively developed, and deploy from a container.
| Stars | License | Stack | Best at | |
|---|---|---|---|---|
| Umami | 38.1k | MIT | Next.js + Postgres | Lightest to run |
| Plausible | 28.4k | AGPL-3.0 | Elixir + Postgres + ClickHouse | The GA-replacement dashboard |
| Matomo | 21.7k | GPL-3.0 | PHP + MySQL | Feature depth |
Umami
The simplest of the three, in the best sense. One page per site, the numbers you actually check (visitors, pages, referrers, countries, devices), custom events when you need them, and a UI that a non-analyst reads without training. It is a Next.js application backed by Postgres and nothing else, which makes it the cheapest to keep alive: the whole stack runs in about 1.5 GB of RAM.
MIT-licensed, which matters if you are embedding analytics into a product you sell.
Plausible
The Community Edition is the closest thing to a drop-in Google Analytics replacement for a content site or SaaS: goals, funnels, UTM breakdowns, and a one-page dashboard that became the visual template half this category now copies. Designed from the start to work without cookies.
The cost of those real-time aggregations is ClickHouse next to Postgres, which makes it the heaviest of the three to run - budget around 4 GB of RAM for the stack - and the piece to watch as traffic grows, because ClickHouse eats disk in a way a Postgres-only tool does not.
Matomo
The old guard, and the only one that competes with Google Analytics on feature count rather than against it: heatmaps, session recordings, A/B testing, funnels (some of these as paid plugins on the self-hosted edition). It is also the only one with a maintained importer for existing Google Analytics data, which matters if leaving GA must not mean starting your history from zero.
PHP and MySQL underneath. It predates the cookieless wave, so out of the box it uses cookies; it can be configured to run without them, at some cost to accuracy. If your instinct is "GA, but mine", this is the one to evaluate first.
What they cost to run
| App | Storage | Total RAM | |
|---|---|---|---|
| Umami | Next.js | Postgres | ~1.5 GB |
| Plausible CE | Elixir | Postgres + ClickHouse | ~4 GB |
| Matomo | PHP | MySQL | ~2 GB |
Google Analytics is free at almost any scale, so the comparison is not price against price. It is a small server plus your operational attention against a consent banner, an opaque event model, and your visitors' data living in an advertising company. On a usage-priced host each of these is a separate line item; on a fixed-capacity plan they run inside capacity you already bought, which is what makes running analytics next to your existing apps effectively free.
Umami and Plausible are both in our catalogue as open-source compose templates: Umami and Plausible. The compose files are vanilla and run anywhere. Matomo we do not ship a template for yet; it deploys fine from its official image.
One operational note that applies to all three: the database now holds your only copy of your traffic history. That is the point, and it is also a backup obligation you did not have when Google kept it.
Do you still need the cookie banner?
Shorter answer than you would like: it depends on what you deploy and where you operate.
Umami and Plausible are designed to work without cookies and without tracking individuals across sites, and that design is the basis on which many teams run them without an analytics consent prompt. Matomo can be configured similarly. But consent requirements are jurisdiction-specific, they cover more than cookies, and a blog post is not the place your compliance posture should come from. What self-hosting genuinely changes is simpler: there is no third party, no cross-site profile, and no data transfer to reason about.
The part self-hosting does not replace
If you run Google Ads, Google Analytics is not just a dashboard. It is the pipe that feeds conversions back into bidding, and the audiences you retarget with. None of the self-hosted tools plug into that, because the integration is the product Google is giving the free analytics away to power.
The honest patterns we see:
- No paid acquisition: self-host and delete the GA tag. Nothing is lost.
- Google Ads matters: keep GA (or server-side conversion upload) for the ads pipeline, and run a self-hosted tool as the numbers you actually look at. Two tags, two jobs.
- Somewhere between: try the self-hosted dashboard for a month next to GA. The usual outcome is discovering which one you stopped opening.
When to stay
- Ads is your growth engine. The conversion feedback loop is the thing you would be unplugging.
- You need GA's ecosystem. Looker Studio reports, BigQuery export, an agency that lives in GA - migrating the tool means migrating the people.
- Nobody owns the server. An analytics gap is invisible until the month you needed the data. If no one will notice a stopped container, the hosted tool is the honest choice.
Frequently asked questions
What is the best self-hosted Google Analytics alternative?
Umami if you want the numbers with the least operational weight, Plausible if you want the closest thing to a full GA-replacement dashboard, Matomo if you need feature depth or an import path for your existing GA history.
Do these tools work without cookies?
Umami and Plausible are designed to, and Matomo can be configured to. Whether that removes your consent banner depends on your jurisdiction and what else your site does - that is a question for counsel, not for a comparison post.
Can I import my Google Analytics data?
Into Matomo, yes - it maintains a Google Analytics importer. Umami and Plausible start from zero, which in practice teams handle by running both tools in parallel until the new history is long enough to stand alone.
How much traffic can a self-hosted analytics tool handle?
More than most sites have. The practical ceiling is the database: Postgres carries Umami comfortably into millions of monthly events, and Plausible's ClickHouse exists precisely because aggregating beyond that is what it is built for. Disk growth, not request volume, is the thing to monitor.
Is self-hosted analytics really free?
The software is. The server, the backups, and the person who notices when ingestion stops are not. If those already exist because you self-host other tools, analytics is close to a free addition; if this would be your first self-hosted service, weigh the standing obligation, not the license.
Does removing Google Analytics hurt SEO?
No. Google states that Analytics data is not a ranking input, and switching your analytics tool changes nothing about how your site is crawled or indexed. Search Console, which is the tool that does matter for search visibility, works regardless of what analytics you run.
What to read next
- Coolify vs Dokploy vs Managed - What self-hosting a platform actually costs
- Fixed-Capacity PaaS: Predictable Cloud Billing for Devs - Why the second self-hosted tool is cheaper than the first
- WordPress on Docker Compose - The production details tutorials skip