The short answer

If you run a single site with modest traffic and you do not want to think about servers, a managed analytics subscription will usually save you time and money for the first year or two. If you care about data ownership, want to skip cookie banners, and have even a small appetite for ops, self-hosting becomes cheaper once you pass roughly 50,000 pageviews per month — and it stays cheaper as you grow. The interesting question is not “which is cheaper” but “what does self-hosting actually include, and what does it cost you in attention.”

Why the choice feels harder than it should

Every comparison online tends to land on the same three open-source names: Plausible, Umami, and Matomo. They are all privacy-friendly, they all run on a small VPS, and they all look fine in a screenshot. The differences that matter — what features you actually get when self-hosting, what the running costs look like at your traffic level, and what the accuracy trade-offs are — are usually buried in the fine print.

That matters because as a founder, you are not picking a dashboard. You are picking where your audience data lives, who has access to it, how much you pay per month forever, and how much of your week goes to keeping the lights on.

What “self-hosted analytics” really means

Self-hosting means a small container (usually Docker Compose) running on a VPS or a homelab server. The analytics script loads from your domain, the data lands in a database you control, and no third party ever sees it. Three tools dominate the conversation:

  • Umami — MIT-licensed, runs on PostgreSQL or MySQL, lightweight script, very quick to deploy. Often the first pick when you want something simple that just works.
  • Plausible — AGPL-licensed, runs on PostgreSQL and ClickHouse, cookieless by default, single-page dashboard. The self-host build is similar in effort to Umami but uses a more performance-oriented database.
  • Matomo — GPL-licensed, PHP and MySQL, the most feature-complete of the three. Heatmaps, session recordings, funnels, A/B testing, e-commerce tracking. Also the heaviest to run.

All three projects were updated in the days before publication and all three offer a managed cloud version as well, which gives you a useful escape hatch if self-hosting ever stops making sense.

What self-hosting actually includes (and what it does not)

Here is the part most comparisons skip. For two of the three tools, the self-hosted build does not include everything the hosted product offers. Some advanced features — funnel analysis, revenue attribution, certain integrations — are documented as hosted-only. That is not a deal-breaker, but it is the kind of thing you want to know before you spend a weekend setting things up.

Practical implication: before you commit, open the project’s feature comparison page and check whether the specific reports you care about are in the self-host build or only on the cloud tier. The default dashboard is rarely the whole story.

The cost picture, honestly

Self-hosting is not free. It is “free of subscription fees,” which is a different thing.

At low traffic (under ~20K pageviews/month):

  • A managed Plausible or Umami subscription often runs around $9/month for the entry tier.
  • A small VPS that comfortably runs any of the three can be had for €3–€5/month if you already shop around, plus a managed database if you do not want to run Postgres yourself (often another €5–€15/month).
  • Real cost at this scale is roughly a tie, but self-hosting eats your time.

At medium traffic (~50K–200K pageviews/month):

  • Managed tiers step up sharply. Plausible and Umami both have higher tiers that can reach $30–$60+/month.
  • Self-hosted infrastructure stays mostly flat: a €10–€20 VPS plus a database is usually enough for all three tools at this volume.
  • This is where self-hosting starts winning on cash.

At high traffic (1M+ pageviews/month):

  • Managed analytics can run into the hundreds per month.
  • Self-hosted costs scale with your VPS and storage, which is predictable, but you may need to learn about ClickHouse tuning, ClickHouse specifically for Plausible, or sharding for Umami before you reach it.

A useful rule of thumb: the hosted plan is cheaper until it is not, and the crossover usually sits somewhere between 30K and 80K monthly pageviews depending on the vendor.

Data ownership and the privacy question

Privacy was the headline reason teams moved off Google Analytics in the first place. By 2026 it has stopped being a differentiator — every credible tool, including GA itself, can be configured to be roughly GDPR-compliant. What differs is how much ownership you actually keep.

  • Self-hosted: Data lives in your database, on your server, in your jurisdiction. You can delete it, export it, or move it whenever you want. You can run everything cookieless and skip the consent banner entirely.
  • Managed privacy-focused (Plausible Cloud, Umami Cloud, Simple Analytics, Pirsch, Fathom): The vendor processes data on your behalf, typically EU-hosted, typically under a DPA, but you are trusting them with operational security and uptime.
  • Google Analytics 4: Free, deeply integrated with Google Ads and BigQuery, but multiple European authorities have ruled that standard GA configurations violate GDPR. Mitigations (Consent Mode v2, EU storage) reduce but do not eliminate the risk.

For most indie founders, the decision is not “data on Google vs data on me.” It is “data on a small vendor I trust vs data on me.” Both beat the default GA setup if you have any EU traffic.

Accuracy trade-offs you should know about

Self-hosted privacy-first tools make deliberate choices that show up in your numbers:

  • Bot filtering. Cookieless tools lean on IP- and user-agent-based bot detection. It works, but it is not as aggressive as GA’s.
  • Session definition. A “visit” in Plausible or Umami is often defined differently than GA4’s session. Numbers between tools will not match exactly. That is fine — you want trends, not forensic accuracy.
  • Sampling. GA4 samples data above certain thresholds on the free tier. Self-hosted tools generally do not sample. Your “true” numbers are usually closer to what self-hosting shows.
  • Real-time. Self-hosted dashboards are unsampled and immediate because the database is yours. GA4 “real-time” is not really real-time.

Pick one tool, use it consistently, and read your own trends rather than chasing cross-tool parity.

When managed analytics earns its cost

Managed is the right call if any of these apply:

  • You do not want a second server in your life.
  • You are at a traffic level where the monthly fee is a rounding error.
  • You need specific features that the self-host build does not include.
  • You are already paying a vendor (your hosting provider, your PaaS) and one more line item does not register.
  • Your analytics needs are simple: pageviews, top pages, top sources, and that is about it.

When self-hosting earns its overhead

Self-hosting pays off when:

  • You run more than one site, or you might.
  • You are above the ~50K pageviews/month crossover.
  • You handle EU traffic and the cookie banner is costing you 40–60% rejection rates.
  • You want full SQL access to your own event data.
  • You already maintain other small services (a blog, a status page, a Matomo-on-Postgres instance) and one more container is not a new category of work.

A practical decision path

  1. List what you actually look at. “Top pages and referrers” is a different decision than “funnels and retention.”
  2. Estimate your 12-month traffic. Be honest — if you are under 20K/month and not growing, the math favors managed.
  3. Decide whether you want a banner. Cookieless self-host = no banner. Cookieless managed = usually no banner. GA4 = banner.
  4. Pick the lightest tool that covers your list. Umami for plain pageview analytics, Plausible for the same with a more opinionated dashboard, Matomo if you genuinely need the full GA replacement feature set.
  5. Try the hosted tier first. Most of these vendors offer a free trial. Run it for a week and confirm the features you care about are present.
  6. Then decide whether to migrate to self-host. Migration is a one-time cost; ongoing maintenance is forever.

What migration looks like in practice

If you are coming from GA4, expect the first two weeks to feel slightly disorienting. Definitions differ, dashboards differ, and your old muscle memory will fight you. The mitigation is to commit to one tool for a quarter before you judge it. After a month the trends are usually clearer than GA4 ever made them, because there is no sampling and no consent rejection noise.

Export your historical GA4 data before you cut over. BigQuery export is free for 14 months; do not skip it.

FAQ

Is self-hosted analytics actually GDPR-compliant out of the box? Cookieless self-hosted tools running on EU infrastructure generally do not require a consent banner. Confirm with your own legal counsel if you are in a regulated industry — this is general guidance, not legal advice.

Can I run more than one analytics tool on the same site? Yes, but each one loads a script. Beyond two tools, you start paying in Core Web Vitals. Pick one for decisions and keep GA4 running in the background only if you have a paid Ads integration that depends on it.

Do I lose features if I self-host? Sometimes, yes. Several advanced features in Plausible and Umami are documented as cloud-only. Matomo’s self-host build is closer to feature-parity with its cloud offering. Check the project’s feature matrix before you commit.

How long does setup take? Umami on a small VPS with Docker Compose is commonly a single afternoon. Plausible is similar. Matomo is more involved — PHP, MySQL, and a heavier footprint — and realistically takes a weekend if it is your first time.

What happens if my VPS dies? You lose analytics until you restore it. Keep automated database backups (most managed database providers include this; on a raw VPS, set up daily pgdump to object storage). This is the single most common self-hosting failure mode.

Sources