The word covers at least five distinct activities, and a company that has bought a product for one of them and is being judged on another has a problem that no amount of configuration will fix. Naming which one you are doing is most of the way to knowing what to buy and who should run it.
Explicit reference: the base
Written answers to recurring questions, searchable, owned and dated. This is the kind with the clearest test of success and the cheapest software, and it is what most companies mean when they say the word. If you are here, buy a base and start from the question list. The software question is nearly settled here, which is why arguing about it is a way of avoiding the writing. What is not settled is who owns each article and what happens when its review date passes, and a company that answers those two before it shortlists a product will be happy with almost any product it picks.
Tacit transfer: getting judgement out of people
Apprenticeship, pairing, shadowing, structured debriefs. No product does this and every product claims to. What software can contribute is the factual residue: the debrief note, the decision record, the list of exceptions that came up. Expecting more is the standard disappointment. The one mechanism that reliably moves judgement is putting a less experienced person next to a more experienced one on real work and then asking the pair, afterwards, what the decision turned on. Write down the answer to that question rather than the whole case, because the case is not repeatable and the reasoning sometimes is.
Lessons learned: what the last project taught
Distinct from the base because it is organized by event rather than by question, and because nobody searches it spontaneously. It works only where somebody is required to read the relevant entries at the start of the next project, which is an organizational rule and not a feature. The failure mode is a register that grows and is never read, and it is not solved by better software. It is solved by naming a moment in the next project when somebody has to read the entries tagged with the same risk, and by keeping the entries short enough that reading them costs ten minutes rather than an afternoon.
Expertise location: who knows this
A directory of people and what they can answer. Small companies do not need it and large ones cannot keep it current without tying it to something people update for their own reasons. Standalone skills databases are among the most reliably abandoned systems in this category. Where it does work, it works because the data is a side effect of something else people maintain for their own reasons: the person who answered this class of question in the support queue, the person whose name is on the article, the person who reviewed the last change to that system. Derived directories stay current; declared ones do not.
Communities: the place people ask each other
A forum or a channel where the answer is produced by conversation. It is genuinely useful and it is not a base: the answer is buried in a thread, dated, and undiscoverable a year later. The productive relationship is that the community produces questions and somebody promotes the recurring ones into articles. Give somebody the job of watching for the same question appearing a third time and turning it into an article, and the channel becomes an intake for the base rather than a competitor to it. Without that job the two surfaces drift apart and people learn that asking is faster than searching, which it is, for them, and expensive for everybody else.
Questions people ask about types of knowledge management
Can one system cover several of these?
A base plus a channel covers the first and the last, and that pairing is what most companies actually need. The other three need organizational rules more than they need software.
Which type is a wiki?
The first, with weaker ownership. A wiki is a base whose governance is by convention rather than by feature, which works well in engineering cultures and poorly almost everywhere else.
Where do tacit and explicit knowledge come from as terms?
They come from the philosophy of knowledge and they are useful as a distinction and misleading as a plan, because the interesting content is rarely purely one or the other. Judge by whether it can be written as an answer somebody else can follow.
Which type should we start with?
Reference, in almost every case. It is the only one with a same-quarter payoff, and the discipline it builds, owners and review dates, is what the others will need later.
Do we need a separate lessons learned system?
Not a separate product. A category in the base and a rule that the next project reads it beats a second system that has its own login and its own decay curve.
Is document management one of these types?
It is a neighbor rather than a type. Document management is about the custody and retention of files; these five are about somebody being able to act on what is inside them.