$ decide --host
When self-hosting actually pays
The self-hosting argument is normally made on licence cost alone, which is the one number guaranteed to favour self-hosting. Here is the comparison with the awkward parts included.
Updated 2026-09-10
The short answer
Self-hosting saves money when the SaaS is priced per user and you have many users, and loses money when the tool is business-critical and you have few. The break-even usually arrives somewhere around 20 to 30 seats on a $10-per-user-per-month tool - but only if you count your own operational time, which is where most self-hosting calculations quietly cheat.
Licence cost alone always favours self-hosting. Adding two hours of operational time a month at $60/hr moves the break-even to somewhere around 20-30 seats - which is where most of these arguments quietly cheat.
- SaaS licence
- Self-host: infra + your time
| 10 seats | SaaS: 100 $/month | Self-host: 160 $/month | SaaS cheaper |
|---|---|---|---|
| 25 seats | SaaS: 250 $/month | Self-host: 190 $/month | break-even zone |
| 50 seats | SaaS: 500 $/month | Self-host: 220 $/month | self-host cheaper |
| 100 seats | SaaS: 1000 $/month | Self-host: 280 $/month | clear win |
The costs a self-hosting calculation usually omits
- Patching and upgrades, forever, including the upgrade that breaks something
- Backups, and actually testing a restore
- Being on call for it - if it breaks on a Saturday, who fixes it?
- Onboarding the next person who has to understand your setup
- The migration cost when you eventually move off it anyway
Good candidates for self-hosting
- Tools priced per user where the user count is large and the tool is not business-critical
- Anything where data locality is a genuine requirement rather than a preference
- Services with a mature container image and a boring upgrade path
- Internal tools where an hour of downtime is an inconvenience, not an incident
Bad candidates, however tempting the price
- Email. Almost always a mistake now - deliverability alone will consume more time than it saves
- Anything customer-facing where downtime costs revenue directly
- Identity and single sign-on, unless you genuinely have the expertise in-house
- Anything with a compliance obligation you would then have to evidence yourself
Common questions
Is self-hosting more secure because the data is ours?
Only if you patch faster than the vendor does, which most small teams do not. Control and security are different things. Self-hosting gives you control, and hands you the responsibility that came with it.
What is the single best thing to self-host first?
Something internal, non-critical and container-shaped, where a Saturday outage is merely annoying. That teaches you the operational reality before anything important depends on your judgement.