Self-Hosted vs SaaS: When It Is Actually Worth It

The self-hosting argument usually goes like this: five SaaS subscriptions at $20 a month is $1,200 a year, and a $5 VPS can run all five, so self-host and save $1,140.

That maths is wrong in a specific and predictable way. It prices the server correctly and prices your time at zero. Once you price your time, the answer changes depending on who you are, and that is the only honest way to decide.

What self-hosting actually costs

The server is the smallest part. The real bill has four lines.

Setup. A Docker container is easy. A Docker container with persistent storage, a reverse proxy, TLS, backups, and monitoring is a day. Multiply by the number of services you are running.

Updates. Every self-hosted service ships security patches. If you do not apply them, you are running known-vulnerable software on the public internet, which is worse than the SaaS you were avoiding. If you do apply them, something breaks roughly one time in ten, and you find out by using the service.

Failure. When a SaaS has an outage, you wait. When your self-hosted instance has an outage, you are the on-call engineer, at whatever hour it happens, including the hour you are trying to give a presentation.

Backups that actually restore. Everyone has backups. Far fewer have tested a restore. An untested backup is a hope.

Add those up for five services and you are looking at a few days a year in steady state, more in the first year. At a freelance rate of $50 an hour, a few days is more than $1,200.

When the maths genuinely works

Self-hosting wins in four situations, and they are worth naming precisely because the general advice "it depends" is useless.

You already run infrastructure. If you have a server, a deployment pipeline, and a monitoring setup for other reasons, adding a service is incremental work measured in minutes. The fixed costs are already paid.

The data is genuinely sensitive. Health records, legal files, customer PII under a contract that forbids third-party processing. Here the question is not cost, it is whether you are allowed to use SaaS at all.

Your usage is far outside normal tiers. A service that charges per seat becomes absurd at a certain headcount. Metabase versus a BI contract, or Umami versus an analytics plan priced by pageviews, can be a ten-fold difference.

You are learning, and the learning is the point. Running a service teaches you things about deployment that reading about it does not. That is a legitimate reason, as long as you are honest that it is the reason.

When it does not

You are a solo founder with a product to ship. Your scarce resource is attention. Every hour on a backup script is an hour not spent on the thing that generates revenue. Pay the subscription.

The service is on the critical path for other people. If a team depends on the tool, your self-hosted instance is now a team-wide risk with a single point of failure who might be on holiday.

You have never restored a backup. This is not a skill you want to acquire during an incident.

The SaaS has a real free tier. Self-hosting Bitwarden to avoid paying for a family plan is a bad trade. Self-hosting it because you want to control the server is a different and reasonable trade.

The hybrid that most people should use

The useful distinction is not self-hosted versus SaaS. It is stateful versus stateless.

Stateful services hold data you cannot recreate: a database, a file store, a notes archive. These are the expensive ones to get wrong, and they are where a managed service earns its price, because backups and durability are the hard part.

Stateless services can be rebuilt from a config file in an hour: a dashboard, an automation runner, an analytics collector. These are cheap to self-host and cheap to abandon.

So: managed database, self-hosted everything else. That is the arrangement that captures most of the savings with a fraction of the risk.

A concrete example

Take a small team paying for analytics, BI, and automation.

Total SaaS: roughly $150 a month. Total self-hosted: roughly $5 a month plus a day of setup plus the ongoing maintenance. Break-even is a few months. After that you are trading your attention for money, at a rate you should calculate rather than assume.

The question to ask

Not "can I self-host this" — the answer is almost always yes. The question is: if this breaks at 11pm on a Friday, is the person who fixes it me, and am I okay with that?

If yes, self-host. If the honest answer is no, the subscription is not a cost, it is insurance, and it is priced fairly.