The comparison people make is a monthly subscription against nothing, and on that comparison the open source option always wins. The comparison that matters is the subscription against the four jobs the hosted vendor was doing for you, priced at what an hour of your own engineering actually costs. Sometimes it still wins. Often it does not, and the reason is never the software.
Job one: it has to stay up
A base nobody can reach during an incident is worse than a base that does not exist, because people planned on it. Self-hosting means somebody owns the host, the backups and the restore test, and the restore test is the part that gets skipped until the day it matters.
Job two: it has to stay patched
A public-facing customer base on unpatched software is an incident waiting for a scanner to find it. The patch cadence of whichever project you pick becomes your cadence, and somebody has to read the release notes and decide whether this one is urgent.
Job three: search has to be good enough
This is where the free options separate. Some ship a full text index that is genuinely adequate for a few hundred articles; others expect you to stand up a separate search service and wire it in. Search quality is most of what a reader experiences, so a project that defers it has deferred the product.
Job four: somebody has to say no to customisation
The freedom to change anything is the trap. Every local modification is a thing that has to be re-applied at the next upgrade, and a base three versions behind because the upgrade is painful is the standard ending. Companies that succeed with open source here usually run it close to stock and put their effort into the writing.
When it is clearly the right call
When the content cannot leave your own infrastructure, when you have an existing platform team who will treat it as one more service rather than a favor, or when you are large enough that the per-seat bill has passed the cost of a fraction of an engineer. Outside those cases the hosted option usually costs less all in.
Questions people ask about knowledge base software open source
Which licence should we care about?
Read whether the license is one the Open Source Initiative recognizes, then read what it asks of you if you modify and redistribute. For an internal base you are usually not redistributing, which makes most of the copyleft argument moot.
Is a wiki engine a knowledge base?
It can be made into one. What it usually lacks out of the box is per-article ownership and review dates, which are the two features that keep a base true. Adding them by convention works until the person enforcing the convention leaves.
Can we start open source and move later?
Yes, and the thing to protect is the export. Before you write two hundred articles, export ten and look at what comes out. If it is clean markup with the titles and categories intact, a later move is a weekend; if it is a database dump, it is a project.
Does open source mean no cost at all?
It means no license cost. Hosting, backups, the search service if you need one and the hours to run it are all real, and a company that has never priced them tends to discover them in the second year.
What about hosted versions of open source products?
Often the best of the trade: the vendor runs it, you keep the export and the option to take it in house. The price is usually below the closed competitors and above zero, which is the honest number anyway.