Confluence as a Knowledge System is like Democracy for Political systems: the worst ever, except for everything else that was tried. As much as I'd like to replace it, it did everything we needed and was better than any alternatives that we considered for both tech/non-tech teams.

The solution we had was to give one Space for each team/department in the company: one space for Customer Support, one for Engineering, Data Scientists, Business Analysts and Operations/Management. Each team lead was responsible in ensuring that other teams from the company could find the information needed, but aside from that we let each team organize itself.

The only hard rule was the one I stated above, everything else came somewhat organically. Managers and Business Analysts, for instance, still preferred to work in Word/Excel, so their pages was mostly just placeholders for Office Docs and attachments. This was fine by me, as long as they could always respond quickly with the right file/link if someone else in the team asked for something.

> The solution we had was to give one Space for each team/department in the company

This is the single biggest tip I would give any company who uses Confluence. For some reason, most companies I work for that use Confluence insist on using minimal separation of Spaces. I.e. IT in general has a Space, and my team has a folder nested 6 folders deep in some path that I can never remember and have to bookmark.

It makes search effectively useless. If I search for "X api documentation" I get 10,000 results ranging from meeting minutes that some unrelated team put up that just happen to use those terms, to pages from other teams documenting how they use the API.

A Space should be treated the same as a Jira project. If I had my way, I would create a Space every time we created a Jira project.

And my biggest request for the Confluence team would be better markdown support (I should be able to indicate that I want to type Markdown without having to click through a giant list), and a way to set up a space to sync from Markdown files stored in Git. Some of us don't want to use the WYSIWYG editor, and I would love to be able to use an MR review process for documentation changes before they go live.

From Steve Yegge's famous Rant about Google not understanding platforms:

    So one day Jeff Bezos issued a mandate. He's doing
    that all the time, of course, and people scramble like
    ants being pounded with a rubber mallet whenever it 
    happens. But on one occasion -- back around 2002 I 
    think, plus or minus a year -- he issued a mandate that
    was so out there, so huge and eye-bulgingly ponderous, 
    that it made all of his other mandates look like 
    unsolicited peer bonuses.

    His Big Mandate went something along these lines:

    All teams will henceforth expose their data and
    functionality through service interfaces.

    Teams must communicate with each other through these
    interfaces.

    There will be no other form of interprocess
    communication allowed: no direct linking, no direct     
    reads of another team's data store, no shared-memory 
    model, no back-doors whatsoever. The only communication 
    allowed is via service interface calls over the 
    network.

    It doesn't matter what technology they use. HTTP,
    Corba, Pubsub, custom protocols -- doesn't matter. 
    Bezos doesn't care.

    All service interfaces, without exception, must be
    designed from the ground up to be externalizable.
    That is to say, the team must plan and design to be 
    able to expose the interface to developers in the 
    outside world. No exceptions.

   Anyone who doesn't do this will be fired.

   Thank you; have a nice day!

I think about that a lot when designing any system, machine- or human-centered, that needs to communicate with a disparate group of entities. Encapsulate the implementation details to those who are closest to the actual information and have enough power to take action, and make the contact interfaces very explicit. This alone will get you 85% of a functional system. The rest is just selecting actors in your system that know how to cooperate and understands these boundaries.

The biggest problem is that "everything else that was tried" wasn't from experts who truly understand information management, nor has our industry learned patterns from other industries that also have similar problems.

I'm trying to do that work now. It's not even that I don't think it can be done in confluence or wikis in general, but rather we don't have a great methodology in organizing it.

I'm keeping a list of what I find at https://www.ebiester.com/annotated-links/documentation/2020/... but I'm not seeing what I really want just yet.

I am afraid I won't be able to help you much there. Every place I saw trying to come up with a grand top-down scheme to organize documentation failed.

Too much structure, and people will wait until they have all of the details needed to start contributing, or they will not contribute unless it's "their job". Too much flexibility and they don't know where to start.

I found that it was always better to let documentation and knowledge sharing follow the same principles as code: let it come bottom-up, see what patterns emerge and only worry about structure when the current practices hit a bottleneck. But then again I never worked on any project that I had hundreds of people directly depending on the documentation, so it might not work as a strategy for larger-scale organizations.

DokuWiki works miles better for me than Confluence ever did.

I've thought about this problem a lot and the conclusion I came to is that tools like Confluence are like the original Yahoo directory of links. Organizing seems like a good idea when there are a small number of items, but it quickly becomes unwieldy. What I've settled on, while not perfect, is shared Google docs where owners of the doc allow suggestions to help keep it up to date.

Also have similar experience with Confluence, as an alternative: we adopted a VuePress static site that we host internally (behind proxy), which basically just renders a bunch of markdown files. For easy discoverability of content we use Algolia's documentation search on top of it. For updating content we either use git locally for technical people, non-technical people can also directly update the content on remote git via gitlab interface. This has been working really well for us once setup, the main challenge we still have now is how to effectively communicate changes to the relevant people at the most appropriate time.

Your knowledge system keeping in sync with your version control is also a major plus, this is the primary reason why I don't like tools like Confluence, because it's easy for them to get out of sync with the actual system, and this can require too much coordination.

> it's easy for them to get out of sync with the actual system, and this can require too much coordination.

I don't follow. How much do you have of your non-techies updating your knowledge base and how do they get to know if what they are updating is "in sync" with the system if the changes are making are orthogonal with the functionality?

It seems your case is mostly about documentation of the technical product, specs, requirements. Yeah, this can go to your version control, and I would bet that you could find a Confluence plugin that would generate the documentation from version control and sync with the rest of your Confluence pages if you wanted.

But I really, really doubt your method can scale beyond the IT side of any middle-to-large organization. The moment you ask Business/Customer Support people to actually go look at anything like Gitlab/Github's interface is the moment you lose 90% percent of buy-in from both non-techies and management at any company that is not absolutely dominated by Software Engineers.

Requiring integration with version control is a convenience for engineers, but it is so far from the tooling from those that live on Excel sheets that you might as well ask them to never touch the documentation side.

Granted, we're still a fairly small organization (+- 10), but our business and customer support people are also using Gitlab to maintain our knowledge base: once you properly guide them on how to use it, they are definitely able to handle this. All the actual git operations happen under the hood of the web interface. So we do strongly believe this is scalable, at least for our use case. This has been working fine in fact for long-lived information across multiple departments (not only technical docs, but also e.g. sales/customer support guidelines). For short-lived information we use Google Docs (e.g. meeting summaries).

> All the actual git operations happen under the hood of the web interface.

The interface is not so much what I object to it. It is the fact that you mentioned "being in sync with the system", which I assume you mean "being in sync with the codebase/released code".

There are plenty of things that happen on any company that have no relation to the code, and you are telling me that you have people on the non-tech side which are expected to understand on what branch of the repo their contributation with documentation is supposed to go? Are their milestones and projects always dependent on the codebase? This is the part that makes no sense to me.

> So we do strongly believe this is scalable, at least for our use case.

I am not telling you that what you are doing is wrong, but scalable it is not. Beware the usage of the Royal We when you say "our use case": are you speaking on behalf on the whole company or just the tech team? This is key difference that you will learn as the organization grows.

> There are plenty of things that happen on any company that have no relation to the code

Some more details about our approach that addresses your concerns: (1) for content changes unrelated to code, anyone (tech or non-tech) is able to open a new MR, and after approval it automatically gets merged into our reference branch and deployed (2) for content changes related to new (unreleased) code, the input from non-tech people (e.g. customer support) is usually requested via an already existing merge request (via a mention), they can then add their input/content via gitlab's interface straight into the markdown files (and see a live preview), this way we ensure this sort of content goes live along with the code and requires no extra coordination

> I am not telling you that what you are doing is wrong, but scalable it is not.

I hope with my clarification above it makes more sense.

> when you say "our use case": are you speaking on behalf on the whole company or just the tech team?

On behalf of the whole company: we have a single product/platform that is continuously increasing in complexity, and it really helps to have a single process in order to maintain a knowledge base that covers as many aspects as possible (tech and non-tech). We believe this approach will also minimize replication of content across departments (facilitates cross-linking), and we're intending to also create an aggregated search (next to the individual search) across all categories of our knowledge base (product/engineering/support/sales/..), in order to more easily retrieve relevant content.

> Beware the usage of the Royal We when you say "our use case"

I am not implying this is set in stone, we (as a whole team) thought long about this flow, and for now it seems to work great, and I hope (responsible for maintaining this technically) that it scales up to some point far enough in the future. I'd say we were mostly inspired by Gitlab, they seem to adopt a similar approach, and that seems to be scaling quite okay over there :). We've noticed ourselves that the general "documentation mindset" has definitely grown among the team because of this specific process and the ease of adding content, updating content and retrieving content.

So essentially you are saying that the tech side still works as gatekeepers of the documentation - need to approve MR's, requesting input via mentions - which is fine until the organization is skewed to the tech side.

As for Gitlab, their work processes are interesting, but I assume that most of the non-technical work there goes to issue tracker and the wiki? I am failing to see where there a strong connection to the code repository is required there.

Afaik, Gitlab's wiki/knowledge base is also a separate repository and content is added/updated via regular MR flow. Our knowledge base is still part of the main codebase due to practical reasons (CI/CD related), but we'll eventually also migrate this to a separate repository as well, to avoid the unintentional gatekeeper aspect as you mention ;).