Self-Hosted Analytics: When It's Worth It (and When It's a Trap)

Self-hosted analytics sound like the obvious move. You skip the SaaS bill, the data sits in your own database, and you can tell investors you “own your stack.” Then six weeks later you are restarting a Docker container at 2 a.m. because Postgres ran out of disk during a traffic spike.

This post is a practical guide to self-hosted analytics for founders and indie hackers. When self-hosting actually pays off, the hidden costs nobody mentions, and when a lightweight managed alternative is the saner call.

What “self-hosted analytics” actually means

Self-hosted analytics means you run the analytics service on infrastructure you control: a VPS, a Kubernetes cluster, or a container on Fly.io. You own the database, the ingest pipeline, and the dashboard. Popular options include PostHog, Plausible, Umami, Matomo, and Rybbit.

The pitch is always the same: no vendor lock-in, no per-event pricing, full data ownership, GDPR-friendly by default. All of that is true. The catch is that “self-hosted” hides a lot of work that a SaaS price tag normally absorbs.

The three real costs of self-hosting

1. Time you do not have

A founder’s most expensive resource is the next three hours. Self-hosting analytics is rarely a one-time install. It is:

• Initial setup (1 to 4 hours if everything goes well)
• Reverse proxy, TLS, domain, CORS (1 to 2 hours)
• Backups and a tested restore path (2 to 4 hours, usually skipped)
• Upgrades every 2 to 6 weeks (30 minutes each, sometimes 4 hours when a migration breaks)
• Disk, CPU, and ClickHouse tuning when traffic grows
• One real outage per quarter, on average

If you are pre-product-market-fit, those hours come out of shipping the product. That is the actual price.

2. Database operations

PostHog uses ClickHouse and Postgres. Plausible uses ClickHouse and Postgres. Umami uses Postgres or MySQL. The dashboards are easy. The data layer is not.

If you have never tuned ClickHouse, your first big merge or a rogue retention query will eat all your RAM and the box will OOM. You can avoid this. You just have to learn ClickHouse first.

3. Compliance you still have to do

People self-host “for GDPR” and then forget that GDPR is about how you process data, not where the bytes physically live. You still need a cookie banner if you fingerprint users, a data processing record, a retention policy, and a way to honor deletion requests. Self-hosting helps with sub-processor disclosure. It does not give you compliance for free.

When self-hosting is the right call

Self-host when you have at least two of these:

• A dedicated person (not the founder) who already runs production infrastructure
• A real compliance requirement that prohibits third-party processors (regulated industries, government)
• Volume above ~5 million events per month, where SaaS pricing starts to bite
• A privacy-first product where “we host your analytics ourselves” is part of the pitch

If you check two or more, self-hosting Plausible or PostHog is reasonable. If you check zero, you are doing it for vibes.

When self-hosting is a trap

Most pre-revenue and early-revenue founders fall here. Symptoms:

• You are evaluating self-hosted options before you have 100 weekly active users
• You picked self-hosted because the SaaS plan looked “expensive” at $29 a month
• You think you will save money on a $5 VPS (you will not, once backups and your time are counted)
• Your team is one person and that person is also writing the product

In all of these cases, the cheapest tool is the one that disappears from your todo list.

A simpler middle ground

There is a third option founders forget: a lightweight managed analytics tool that gives you most of what self-hosting promises, without the ops surface.

What “lightweight” means in practice:

• One snippet, one event API, no SDK zoo
• A free tier that covers a real product, not 1,000 events
• EU hosting, GDPR-friendly defaults, no fingerprinting
• A dashboard you can read in 10 seconds, not a configuration wizard

That is the trade Eventraze is built for. You get a managed Amplitude alternative with EU hosting and a free hobby plan, and you do not run any infrastructure. If you outgrow it later, your event schema is portable: the 7 events worth tracking are the same events any other tool ingests.

The decision in one paragraph

If self-hosting analytics is the most interesting part of your week, you are probably not building the right thing. Pick the tool that gets out of the way. If, in nine months, you have real volume, real compliance pressure, and a real platform engineer, then move to self-hosted. Until then, ship the product.