After Schrems II, several European Data Protection Authorities found that Google Analytics created unlawful EU-to-U.S. data-transfer issues under the old transfer setup.
The EU-U.S. Data Privacy Framework has since changed the legal basis for certified U.S. providers, but many privacy-first site owners still avoid GA4 because it remains tied to Google's ad ecosystem, consent burden, and cross-border processing questions.
For sites that want a cleaner privacy setup, the practical answer is to move analytics off GA4 and onto something self-hosted, in an EU datacenter, with first-party tracking and fewer cross-border processing questions.
Four options keep appearing in roundups: Umami, Matomo, Fathom Lite, and Ackee. Two are alive and worth your time. One is feature-frozen and low-release. One is alive but niche. This post compares all four, gives you a decision rule by archetype, and matches each tool to the VPS plan that runs it well.
TL;DR
The short version is not complicated. Most sites should start with Umami, move to Matomo only for heavier reporting needs, skip Fathom Lite, and treat Ackee as a niche preference pick.
- Use Umami for personal blogs, indie SaaS, and anything where the analytics dashboard needs to be readable by non-engineers. Runs comfortably on a 1 GB RAM VPS. MIT licensed. Cookieless by default.
- Use Matomo when you genuinely need funnels, ecommerce tracking, heatmaps, or session recording in one tool. The full feature surface costs more in RAM and operational attention. Realistic floor is 2 GB RAM for small sites; 4 GB for moderate-traffic sites with default features enabled.
- Do not use Fathom Lite. The open-source Docker image has not been updated in five years. The maintainers have moved fully to commercial Fathom Analytics. Running it as an internet-exposed service in 2026 is a security and compatibility liability.
- Ackee is alive but lower-velocity than Umami. Pick it only if the dashboard aesthetic matters to you specifically. Otherwise default to Umami.
The Four Contenders, Briefly
The shortlist is smaller than it looks. Umami and Matomo are the serious choices for most self-hosted analytics setups. Fathom Lite still appears in older lists, but the open-source version has fallen too far behind. Ackee still works, but it now sits closer to a preference pick than a default recommendation.
Umami
Umami is a Node.js application backed by PostgreSQL. MIT licensed. Active development with weekly commits and regular tagged releases throughout 2025 and into 2026. The current stable release is v3.1.0.
The Node process idles at around 200 MB of RAM. Postgres adds another 150-250 MB, depending on configuration. The full stack runs comfortably on 1 GB total. Tracker script is roughly 2 KB. Cookieless by default; no third-party identifiers; no consent banner needed under most EU regulator interpretations.
- Headline strength: cleanest dashboard of any tool in this set, lowest setup friction, lowest operational footprint.
- Headline weakness: the funnel and heatmap features are thinner than Matomo's, and the multi-site management UI works but is less mature at agency scale.
Matomo
Matomo is a 15-year-old PHP application backed by MySQL or MariaDB. The project was formerly called Piwik. GPL v3 for the core; some advanced plugins (heatmaps, session recording, A/B testing) are commercial. Current stable is v5.10.0.
The realistic RAM floor is 1.5-2 GB total: PHP-FPM workers consume 50-80 MB each, MariaDB wants 512 MB to be comfortable, nginx is small but present, and the archiving cron job spikes both CPU and memory periodically.
- Headline strength: largest feature surface of any open-source analytics tool. Funnels, multi-attribution, ecommerce, goals, custom dimensions, segments, marketing attribution, GA4-style reports.
- Headline weakness: the dashboard is dense (some users find it Excel-like), and the operational attention required is real. You will need to know what an archiving cron job is.
Fathom Lite
Fathom Lite is the original open-source Go implementation by Jack Ellis and Paul Jarvis. MIT licensed. The repository still exists on GitHub, and the latest GitHub release is v1.3.1. The Docker image still pulls.
The Docker image on Docker Hub has not been updated in five-plus years. The repository's master branch has had only sporadic, non-releasing commits in that window. The README says the Lite version is no longer getting new features, though the maintainers still describe it as bug-maintained long-term. The actively developed product is commercial Fathom Analytics.
Running a feature-frozen analytics app with a stale public Docker image in 2026 is a liability. Browser tracking-script behavior has evolved. Runtime and dependency updates that active projects ship regularly are not showing up in the public Lite release path.
If you find Fathom Lite recommended in a 2026 roundup, check the Docker image date and release history before treating it as a production pick.
The commercial Fathom Analytics SaaS at usefathom.com is a different product. It is fine if you accept that it is SaaS and not self-hosted. It is not what this post is about.
- Headline strength: tiny binary, small tracker, simple original product idea.
- Headline weakness: feature-frozen open-source codebase, stale Docker image, unclear release path, and no strong reason to deploy it as an internet-exposed analytics service in 2026.
Ackee
Ackee is a Node.js + MongoDB analytics tool. MIT licensed. Current release is v3.6.0. The maintainer ships bug fixes and dependency updates. Major feature work has slowed. The dashboard is minimalist, the tracker is small, and the resource footprint is comparable to Umami's.
Ackee works. The reason it sits in the niche column is that Umami is at least one velocity tier above it on every relevant axis: development pace, ecosystem, third-party documentation, available integrations.
If you already have a strong reason to prefer Ackee (you tested both and liked the dashboard better), use it. Otherwise pick Umami and move on.
- Headline strength: clean dashboard, small tracker, low resource footprint, still maintained.
- Headline weakness: slower development pace than Umami, smaller ecosystem, fewer integrations, and less third-party documentation for people who need help after install.
Decision Framework: Pick by Archetype
The archetype matters more than the feature grid. A personal blog does not need the same analytics stack as a SaaS product or an agency managing client accounts. Start with the dashboard owner, the reporting depth, and the amount of server work you are willing to take on. The tool choice becomes much simpler after that.
Personal Blog or Single-Site Owner
Pick Umami. The setup is one Docker Compose file. The tracker is a one-line script in your <head>. The dashboard is readable on a phone. RAM and storage are trivial concerns at this scale.
There is no second tool needed. If you are leaving GA4 for principled reasons or to remove the cookie banner, Umami closes the case.
Indie SaaS Founder or Small Product Team
Default to Umami. Add Matomo only if you have specifically identified a missing feature.
The most common reason to add Matomo here is funnel reporting across signup and upgrade flows that Umami does not yet do as well. The decision question is whether someone on the team actively wants to operate Matomo's larger surface area.
If the answer is no, Umami, plus a couple of database queries against your product database, will get you 90% of the funnel insight at 10% of the operational cost. If your product is event-heavy and you want native funnel UI, install Matomo. The operational cost is a larger stack: Matomo, a database, a web server layer, optional Redis, more RAM, and an archive job to manage.
Agency Managing 10+ Client Properties
Pick Matomo. Multi-site management is mature. Per-user permissions and per-client login are first-class. The paid heatmap and session-recording plugins are differentiators that agencies actually invoice for.
Umami's multi-site UI works but feels thinner at agency scale: filtering across many sites, assigning client-level access, exporting branded reports. Matomo's UI is dense, but the density buys you the features the work requires.
If the agency has a portfolio of mostly low-traffic sites and the main job is monthly traffic reports rather than conversion-rate optimization, Umami is still a defensible choice and will save infrastructure cost. The decision is about feature ceiling, not about the wrong tool.
Feature-by-Feature Comparison
| Capability | Umami | Matomo | Fathom Lite | Ackee |
|---|---|---|---|---|
| License | MIT | GPL v3 (paid plugins commercial) | MIT | MIT |
| Stack | Node + PostgreSQL | PHP + MySQL or MariaDB | Go + MySQL or SQLite | Node + MongoDB |
| Active development (2026) | Yes, weekly commits | Yes, full-time team | No, ~5 years stale | Yes, low velocity |
| Cookieless by default | Yes | Configurable (flag required) | Yes | Yes |
| Multi-site management | Yes, basic | Yes, mature | Yes, basic | Yes, basic |
| Funnels and goals | Limited | Full | None | None |
| Ecommerce tracking | None | Full | None | None |
| Heatmaps and session recording | Session replay; no native heatmaps | Yes (paid plugin) | None | None |
| Tracker script size | ~2 KB | ~22 KB | ~1 KB | ~2 KB |
| Realistic RAM floor | 1 GB total | 2 GB total | n/a (do not deploy) | 1 GB total |
| Operational complexity | Low | Medium | n/a | Low |
| Recommendation | Default for most | When feature ceiling matters | Avoid | Niche |
VPS Sizing in Plain Numbers
The right tool depends in part on what you are willing to pay for the box that runs it.
| Setup | Realistic RAM | CPU | Storage | Good VPS floor |
|---|---|---|---|---|
| Umami + Postgres, low traffic | 512 MB-1 GB | 1 vCPU | 20-25 GB NVMe | Small VPS |
| Umami + Postgres, growth headroom | 1-2 GB | 1-2 vCPU | 40-60 GB NVMe | Small to mid VPS |
| Ackee + MongoDB | 1 GB | 1 vCPU | 20-25 GB NVMe | Small VPS |
| Matomo + MariaDB, small site | 1.5-2 GB | 1-2 vCPU | 40-60 GB NVMe | Mid VPS |
| Matomo + MariaDB, moderate traffic, default features | 2-4 GB | 2 vCPU | 80-120 GB NVMe | Mid to larger VPS |
| Matomo + MariaDB + heatmap or session recording | 4 GB+ | 2-4 vCPU | 120 GB+ NVMe | Larger VPS |
Why these floors matter:
Umami's small footprint is by design. Umami stores lightweight analytics events in PostgreSQL rather than full web-server request logs, so database growth stays modest for small and medium sites.
The Node process stays under 250 MB even when several thousand visitors hit the site daily. Postgres growth is modest.
A 1 GB VPS with 25 GB of NVMe holds many years of analytics for a typical solopreneur site. I run Umami on a 1 GB VPS in Frankfurt for a small project; latency from Lagos is around 110 ms, the dashboard responds in under 400 ms, and the box has been quiet for months.
Matomo's footprint is larger because the architecture is older and more general. Each PHP-FPM worker consumes 50-80 MB. MariaDB performs poorly with less than 512 MB allocated.
Matomo's recommended hourly archiving cron can spike CPU and memory, so the VPS needs headroom beyond the normal PHP and database footprint. A 1 GB VPS technically runs Matomo for very small installations, but the dashboard feels sluggish, and the OOM risk during archiving is real. The 2 GB floor is not gatekeeping. It is the box where the tool stops fighting you.
Storage growth is rarely the bottleneck. A 60 GB NVMe disk holds many years of Matomo data even with default raw-log retention enabled. If you turn on session recording, plan for an order of magnitude more disk per month and revisit the plan size annually.
Pro Tip
If you are running Umami today on a 1 GB VPS and your traffic is growing, the upgrade path is straightforward. Snapshot the VPS, resize to 2 GB, restart. The extra RAM gives Postgres and the Node process more headroom.
EU Datacenters and the GDPR Angle
The Schrems-II and GA4 transfer issue from the intro is the legal background here. Self-hosting does not magically make analytics compliant, but it does remove the default GA4 pattern: visitor data leaving your site, entering Google's analytics stack, and raising cross-border processing questions.
Self-hosting your analytics inside the EU keeps the stack simpler at the level of jurisdiction. It is not a compliance cheat code. The data is collected by your tracker, sent directly to your server, processed, and stored in the region you picked.
Your VPS provider may still be a processor, so processor terms, security controls, retention rules, and privacy notices still matter. The point is narrower: there is no analytics SaaS provider sitting in the middle, and no default GA4-style transatlantic analytics transfer.
The datacenter choice should follow your visitors and your compliance posture. Frankfurt is the standard pick for German, Austrian, and broader EU audiences. Amsterdam is a natural fit for Benelux traffic.
London works for UK GDPR cases, but it is outside the EU, so it is not the same answer for EU-only data-residency requirements. Zurich works for Swiss audiences and Switzerland-specific privacy needs, but it is also outside the EU.
Pro Tip
Putting a US-headquartered CDN or proxy in front of your analytics endpoint can reintroduce transfer analysis and processor-review work. If the whole point is EU-only analytics handling, terminate TLS directly on the VPS or document the CDN setup carefully.
Setup Mechanics Overview
This is the deployment shape, not a tutorial. Full step-by-step guides are separate posts.
Umami: One Docker Compose file. Two containers: Umami (Node) and PostgreSQL. One environment variable: DATABASE_URL. Default port 3000. Reverse proxy in front for TLS (Caddy is the lowest-friction option; nginx-proxy-manager and Traefik also work). Add the tracker script (one line) to the <head> of every page you want to track. Upgrades are docker compose pull && docker compose up -d.
Matomo: Docker Compose with three to four containers: Matomo (matomo:fpm-alpine), nginx (or Apache) in front of PHP-FPM, MariaDB, and optionally Redis for caching. First-time browser-based config wizard handles database connection, admin user, and first-website setup. Tracker script is a Matomo-generated JS snippet pasted into the <head>. Required: a cron job for archiving. The default option of triggering archiving via URL (?force_archiving=1) works for small sites but produces visibly slower dashboards. Upgrades involve docker compose pull plus a console core:update invocation.
Both: TLS via reverse proxy is the standard pattern. Both projects publish official upgrade guides. Both have working backup recipes (pg_dump for Umami, mariadb-dump for Matomo).
This is where Cloudzy's marketplace actually matters. You can deploy Umami or Matomo on a VPS without starting from a blank server, writing the Compose file yourself, or spending the first hour wiring the base stack together.
The VPS still needs to be the right box. Analytics likes fast disk, predictable RAM, and a region close to the people loading your tracker script. Cloudzy gives you NVMe storage, DDR5 RAM, up to 40 Gbps networking, full root access, dedicated IPv4 and IPv6, 12+ global regions, 99.95% uptime, and a 14-day money-back guarantee.
For Umami, the benefit is speed: launch, attach your domain, put TLS in front, and paste the tracker script.
For Matomo, the benefit is avoiding the blank-server work before you even get to archiving, retention, backups, and tracking settings.
When Self-Hosting Is the Wrong Call
You are a solo non-technical founder with no Linux experience and no time to learn it. The right answer is Plausible Cloud at $9-19 per month. It is GDPR-friendly, the dashboard is excellent, and you do not own a server.
The math on self-hosting only works if your time is cheaper than the SaaS fee or you genuinely want the operational practice. For a non-technical solo founder, the time math does not work.
You need real-time SOC2 or HIPAA-grade audit trails on your analytics processor. Neither open-source Umami nor open-source Matomo deliver that out of the box. You can build the audit posture yourself, but the work is real and the certification process is its own project. Buy compliance-as-a-service for this case.
Your marketing stack requires GA4 or Google Ads attribution that depends on Google's cohort and remarketing pixels. Self-hosted analytics is not the right tool category for AdWords or Meta-Ads optimization.
The conversion data needs to flow back to Google or Meta to retrain the bidding algorithms. Self-hosted analytics replaces the descriptive analytics use case (what happened on my site), not the ad-attribution use case (which ads should I run).
Closing
The decision rule is short. Default to Umami for personal blogs, indie SaaS, and small product teams. Move to Matomo when funnels, ecommerce, or heatmaps are non-negotiable and you have the operational attention to spare.
Skip Fathom Lite. Pick Ackee only if you have already tested it and prefer the dashboard. Run the box in Frankfurt or Amsterdam if EU jurisdiction matters; otherwise run it where your traffic is.
The infrastructure cost for a working self-hosted analytics deployment is between $7 and $30 per month for the VPS. The labor cost is one Docker Compose file and a reverse proxy. Most of the difficulty is in the deciding, not in the doing.
Frequently Asked Questions
What Is the Best Self-Hosted Analytics Tool in 2026?
Umami is the default recommendation for personal blogs, indie SaaS, and small product teams. It is MIT licensed, actively developed, cookieless by default, and runs comfortably on 1 GB RAM. Choose Matomo for funnels, ecommerce, or heatmaps. Avoid Fathom Lite.
Is Fathom Lite Still Maintained?
No. The open-source Docker image has not been updated in five-plus years, and GitHub activity is sporadic and non-releasing. The maintainers now focus on commercial Fathom Analytics. The Lite version should not run as an internet-exposed service in 2026.
Can Self-Hosted Analytics Replace Google Analytics for GDPR Compliance?
Yes, conditionally. Self-hosted Umami or Matomo on an EU VPS keeps analytics data in your chosen region and removes the default GA4 transfer pattern. Do not put a US-based CDN like Cloudflare in front of the analytics endpoint if GDPR posture is the reason.
How Much VPS Do I Need to Run Matomo?
The realistic floor is 2 GB RAM for a small Matomo site. For moderate traffic with default features, plan for 4 GB. Heatmaps and session recording push that higher. The floor exists because PHP-FPM, MariaDB, nginx, and archiving share the box.
Does Umami Need a Cookie Banner in the EU?
Under most EU regulator interpretations, no. Umami does not write a tracking cookie by default. It uses first-party request data and aggregates it server-side. Some jurisdictions may still expect a notice, so check local DPA guidance if compliance is the main reason.
How Does Umami Compare to Plausible Self-Hosted?
Umami's self-hosting story is smoother. Plausible Community Edition exists, but Plausible's commercial focus is Cloud. Umami treats self-hosting as the main distribution path. For self-hosted-first deployment, Umami is the safer pick. For SaaS, Plausible Cloud and Umami Cloud both work.
Which Datacenter Location Is Best for Self-Hosted Analytics in the EU?
Frankfurt is the default for German, Austrian, and broader EU audiences. Amsterdam fits Benelux traffic. London works for UK GDPR, but not EU-only data-residency requirements. Pick the region closest to your visitors unless compliance needs point somewhere more specific.
