On Stack Exchange site Epistemology
I'm going to make an argument that the current Stack Exchange sites are (a) too broad and (b) aligned in ways that may alienate significant communities of potential users. One can observe situations where Stack Overflow does a poor job of getting answers to long-tail questions, probably due to their being lost in the noise. I can also see at least one significant and fairly coherent field where I directly work. Subject material pertinent to that field covers multiple sites and has a significant body of related knowledge where none of the sites appear to be a good fit.
This piece discusses the issue and an example of a field with a brief explanation of the reasoning as to why Stack Exchange sites aren't a particularly good fit for the domain.
Falling in the cracks
I have seen one or two examples of (nominally) development related topics that tend to get poor response off Stack Overflow, which I'll discuss below. Conversely, I can think of topics (one specifically) that sort of fall in the cracks between the sites and even other proposed sites.
On a couple of occasions before, I've suggested that a M:M relationship between sites and questions might not be a bad thing, although it might be a bit of a performance to implement. In this case, moderators or users with appropriate levels of reputation points could poach questions from one board and reflect it on another.
There is also enough overlap between people's perceptions of the boundaries that some questions could sensibly live on different boards - and different moderators may well have different opinions about what belongs where.
Without attempting to plug the proposal, I suggested a site on Area 51 that would have significant overlap with Stack Overflow, Server Fault and Super User, plus some overlap with other proposals on the Area 51 site. In my opinion (which is why I proposed it) the epistemology of the proposal would suit a coherent (and potentially quite large) body of users with a range of questions that might have significant overlap with several other sites.
One could argue that the large user base on Stack Overflow is a bit of a double edged sword. Obscure or highly specialised long tail questions do actually tend to get lost in the noise. While there are people who may well know the answer, the volume of .NET questions tends to drown other things out.
However, I don't see that splitting Stack Overflow into .NET and 'everything else' would work as people aren't going to go to a 'Stack Overflow for non-mainstream stuff that nobody's interested in' board. Splitting or specialising this type of question could only really be done on the basis of topic areas that people can actually relate to. People would actually have to have boards with topic areas that they identify with.
From a 'market segment' or sociological point of view, Stack Overflow style boards would work best if the split went with a topic area that people can relate to. Currently we have 'Programmers', 'System admins', and 'computers for everyone else'. I think any other sites really need to be for things that people could identify themselves with, so they feel some affinity to the site and a reason to 'hang out' there.
Case study (warning: implicit plug for proposal on Area 51)
I put up a proposal for a 'Business Intelligence' forum on Area 51 (not telling you anything you can't see by stalking my account there). There was some specific reasoning behind this, although the site isn't all that good at providing space for supporting blurbs to explain this sort of thing. The main reasons I set up this proposal were:
Business Intelligence is a fairly multi-disciplinary activity. Questions in this field could easily fall into Stack Overflow, Server Fault or Super Users's catchment.
Quite a lot of practitioners in this field don't view themselves as programmers - often people working in this space are power users or semi-technical 'IT-finance' types who wouldn't intuitively identify with a programming forum.
This field actually employs more staff than software development does. If you look at statistics on job boards, far more jobs are tagged with 'SQL' than even major programming languages like Java, C or C#. Most of these positions are actually analyst roles where 'database skills' are required for reporting.
Again, people in these roles are probably not going to identify with a programming forum. I'll argue that this is potentially quite a large pool of punters for a Q&A site that would fall in the cracks between current Stack Exchange offerings. Note, however, that 'getting the word out' to this group of people might be somewhat slower than (say) to an audience of Internet-centric software developers.
- It's not simply an Excel 'power user' forum, as many such people are actually using specialised Business Intelligence tooling (one of the main SQL Server Analysis Tutorials on the web is actually written by an accountant), and a lot of 'tools guys' in the industry are actually self-taught people from the 'business side'.
In most cases (I build data warehouse systems for a living) these people are actually some of the primary customers of the systems I build. Quite often they find themselves building ad-hoc systems - I've seen plenty of instances of multinational companies where the finance department maintained their own data warehouse system, essentially built by 'amateur' developers from within the business.
This field covers quite a wide range of people, levels of technical skill, and tooling ranging from Excel to Business Intelligence tools to bespoke software. Quite a lot of people active in this field would not see themselves as developers, and might well relate more to a 'Business Intelligence' forum than a scattering of Stack Exchange sites.
Enough of the plug, what's the point?
I think this is an example of a field where the current boundaries of the Stack Exchange sites are not really a good fit, but it does have significant overlaps with the existing sites (other examples might be 'ERP Systems' or SEO). I also find that Stack Overflow is farily ineffective at getting answers to deep questions about Business Intelligence tools - for example, try asking technical questions on SQL Server Analysis Services, Cognos or MicroStrategy there.
I also would suspect that quite a few questions the 'power users' end find themselves getting pushed off to Super User, which is also not terribly good at answering 'power user' questions. Around 1.1% of the questions on Super User are tagged 'excel' and 'access' and 'ms-access' make about 0.25% of the questions. Questions about VBA sometimes get referred to Stack Overflow. This suggests that the epistemological boundaries of the sites are a poor fit for this type of user.
Get to the point
From this we can see an example of a field which one might intuitively think is an appropriate fit for Stack Overflow, but in fact is poorly served by the existing arrangement of the sites. There are probably other potential areas of interest to significant bodies of people where the sites in their current form look like they might serve but probably don't in practice.
On this basis, you might well get better results by farming off bits of subject material to more specialised sites. While people with the knowledge to answer to the question might frequent Stack Overflow, my experience is that specialised long-tail questions are quite likely to get lost in the noise.
When Stack Overflow was younger - in its public beta phase - the volume was much smaller, and individual questions were quite likely to get more attention. This suggests that there is actually an optimum community size for this type of forum. Loosely, you could view the probability of getting an answer to a long-tail question as a function that is:
Proportional to the number of users on the site who know about the topic.
Inversely proportional to the number of unrelated questions pushing the question down the 'new' or 'hot rankings, which is proportional to the number of users asking questions on popular topics and the number of people answering questions on popular topics. You can see a network effect here that will strongly favour questions on mainstream topics.
Probably also adversely affected by the likelihood that people who know about your topic aren't interested in the mainstream topics and get turned off by the fact that they never see any questions on topics that they are interested in.
Thus, there is probably an optimal community size and level of coherence for a site that will produce the best results for long tail questions on a given topic. If the sites were something that a coherent body of people could relate to then smaller sites with a more focussed community might actually do a better job of providing answers than an all-encompassing Stack Overflow.
Conclusion
I'll argue here that the current epistemology of the sites is too general, and this arrangement is not doing a good job of answering questions in some long tail fields that are nominally supposed to be covered by the existing sites. To do a really good job of long-tail questions, sites need to be optimised for:
A community that people can identify with, so that punters would see something that they intuitively view as being of interest to them. It would also have to be something that would attract a core of expert users, so it can't be dumbed down too far.
A level of speciality and focus on a coherent subject area, so that you don't see too much of the 'poor cousin' effect observed on more esoteric topics on Stack Overflow.
The right size community to keep the average likelihood of meeting eyeballs at a level where there enough users but not such a large volume of questions that stuff gets missed.
Thus one could argue for a set of more specialised boards with distinct communities, even within topics that are currently nominally covered by other sites. Specifically, topics that are not particularly well served by incumbent sites due to the 'poor cousin' effect are likely to be better candidates.
The epistemological breakdown should not be too granular. SAP R/3 (or any specific product) is too granular. 'ERP & CRM Systems' might be a better fit. I think the optimal breakdown for the purposes of forming a community would probably follow distinct groups of people who would view themselves as being in a professional field (hence 'Business Intelligence' as an example above).
Some thoughts on what might or might not make good communities:
Databases: Maybe, but Stack Overlow and Server Fault do a passably good job of this now so I don't think there's a practical deficiency in the cover being provided already. However, other boards like SQL Server Central cover programming and admin questions. A database specific board (that is, split Server into server/infrastructure and database boards and aggregate database content from Stack Overflow) might work a bit better. However, the existing service provided by Stack Overflow and Server Fault isn't significantly lacking, so this is probably best viewed as a case of 'if it ain't broke, don't fix it.'
Business Intelligence: already discussed at length above.
Content Management systems: Although not explicitly in the brief of Stack Overflow, this is for all intents and purposes a core topic in the Stack Overflow community. Probably counterproductive unless there are a lot of long-tail questions getting missed that I didn't notice.
Kernel hacking: probably too specialised and it tends to be fairly served by existing forums like the LKML.
U.I. design and HCI: not particularly well served by Stack Overflow and the field includes graphic designers, systems analysts, QA and interaction people who might not necessarily relate to a programming community. Maybe a candidate for another board.
Excel/Access power users: Somewhat product specific, but falls in the gap between Stack Overflow and Super User and (in practice) not particularly well served by either. Large numbers of potential users (assuming that they can be made aware of such a board), so it may be viable on this basis alone. There are certainly quite a few incumbent sites on this subject, so there's evidence of demand, if not necessarily an open niche. May be worth doing if the Stack Exchange people decide that they want to be expansionistic.
Other specialised development areas such as game development: If there is evidence of a sufficient body of people who might identify with the subject area and a tendency for long tail questions to get drowned out then fields within programming might be a candidate for their own site. If a specialised board would do a better job of addressing the long tail stuff and it can get a large enough user base you might make a case for segmenting it off Stack Overflow.
