Self-hosted versus hosted (SaaS) is not a question with a single right answer. It turns on how fast your board will grow, whether anyone wants to be paged for a support tool, and whether the self-hosted option even needs a server. Below is the table to run your own situation through, not an argument for either side.
What actually decides self-hosted versus hosted?
Five things decide it, and none of them is "which is cheaper" on its own. Match your own situation against each row before you look at a single price.
| What's true for you | Leans hosted | Leans self-hosted |
|---|---|---|
| Growth | Your active users will likely stay in the low hundreds | You expect real growth in the number of people posting and voting |
| Who's on call | Nobody on the team wants pager duty for a support tool | Someone already owns patching and backups for other services you run |
| Budget shape | A predictable $22 to $29 a month suits you better than a variable bill | $70 to $125 a month, or a $10 to $20 VPS, plus five hours a month, is affordable |
| Compliance | You need SOC 2 or similar evidence now | You have no third-party compliance requirement to satisfy |
| Architecture | You want zero infrastructure of any kind, including a server you don't manage | You're comfortable running Docker and Postgres, or your chosen tool needs neither |
If most of your answers land in the first column, stay hosted. That is the honest answer for most small teams, and it is not a consolation prize: it is one fewer thing to own. For the fuller cost and options breakdown behind this table, see self-hosting a feature request board.
Why do the self-hosting cost estimates disagree by ten times?
Two sources in this category publish real numbers, and they do not agree. UserJot's costed comparison against self-hosting Fider lands on $70 to $125 a month in infrastructure before any engineering time: about $20 for Postgres, $20 to $40 for the host, $10 to $35 for transactional email, $20 to $30 for monitoring and TLS, plus five hours a month once it is running. Quackback, an open-source project arguing the opposite case, puts a small team's cost at $10 to $20 a month on a VPS, rising to $50 to $100 for a larger one.
Both figures are honest. They are pricing two different things. UserJot is costing a managed setup: a separate database service, a separate host, a separate email provider, and separate monitoring, each billed and maintained on its own. Quackback is costing one small server doing all of it at once. That is cheaper on paper, and it comes with a trade the number does not show: the box running your database is very often also the only box you could restore that database onto. A managed setup keeps those two risks apart. A single VPS does not.
Neither number is wrong. Pick the setup first, and the true cost follows from that, not the other way round. The open source options compared shows which projects assume which setup.
When does a hosted subscription start working against you?
Some hosted pricing is flat. Some grows with exactly the thing you want to grow: your users. Canny.io prices by tracked user, defined on its own pricing page as anyone who posts, votes or comments, and that count only rises. Billed yearly, its entry tier is $79 a month at 100 tracked users and $1,079 a month at 5,000. Billed monthly, the same two points are $99 and $1,349. Featurebase prices per seat at $29 a month billed yearly, so three colleagues alone is $1,044 a year before anything else.
Not every hosted tool works this way. Upvoty prices at $15 a month with unlimited tracked users on every tier, which is the opposite model. The question worth putting to a vendor before you sign up is not "what does it cost now" but "what does the price scale on, and what happens if this actually works." What feedback tools cost has the full pricing set this table draws from.
One payment, no subscription, unlimited products.
Is there a middle ground?
There is a third option the table above does not have a row for: a self-hosted tool with nothing to operate. Docket is built as a static site plus two small endpoints, so there is no Postgres to run, no Docker container to keep alive, and no mail server to configure; the free tier of any static host covers the rest. That removes most of the self-hosted column's cost and effort in one move, because there was never a server to patch or a database to back up.
It does not make the hosted column wrong. A team with nobody who wants to touch infrastructure of any kind is still better served paying a vendor $22 to $29 a month and never thinking about it again. The static-site option only matters if you were already leaning self-hosted and the server was the part putting you off.
Frequently asked questions
Is self-hosting always the cheaper option?
No. UserJot's costed comparison puts a managed self-hosted Fider deployment at $70 to $125 a month in infrastructure alone, against $22 to $29 a month for the cheapest hosted subscriptions in this category. It only comes out ahead once you also count what a growing hosted bill would have cost, or when the self-hosted option runs on a single cheap VPS instead of managed services.
Why do the two published self-hosting estimates differ so much?
Because they cost different setups. UserJot's $70 to $125 a month prices a managed database, a managed host, transactional email and monitoring as four separate services. Quackback's $10 to $20 a month prices one VPS doing all of that at once, which is cheaper, but means the same box holds both your data and your ability to restore it.
Which hosted pricing models grow the fastest with success?
Tracked-user and per-seat models. Canny.io's entry tier runs from $79 a month at 100 tracked users to $1,079 a month at 5,000, billed yearly. Featurebase charges $29 per seat per month billed yearly, so cost rises with headcount rather than usage. Not every vendor prices this way: Upvoty charges a flat $15 a month with unlimited tracked users.
What if my team is too small to be on call for anything?
Then the table has already answered it. Stay hosted. Five hours a month of patching, deliverability and backup checks is a real cost even when the infrastructure bill is small, and a team with nobody to absorb that time should not take it on for a support tool.
Does a smaller team always mean staying hosted?
Not automatically. A small team that already runs its own servers for other things, and only needs a static board with no database, can self-host cheaply precisely because there is nothing new to operate. Team size matters less than whether the tool you are choosing needs a server at all.