22

I have a simple question, per the title, that I'd like community feedback on before venturing over to Stack Exchange meta and/or bugging one of the Stack Exchange Community Managers.

Should the rep requirement for creating tags on Server Fault be raised? (And if so, to what?)

Currently, it's at 300 for Server Fault. By contrast, at Stack Overflow, 1500 rep is needed to create new tags. I think our threshold is too low, and believe that the lower threshold here is the reason that we have literally hundreds of garbage tags with only a handful of questions in them. Because I stumbled across them within the last few minutes, , and come to mind immediately as tags with absolutely no value that had one or two questions in them, and were created by users with low reps. (Not to say that all tags with few questions are necessarily garbage tags, but there is a strong correlation, from what I've seen in my retagging efforts.)

I am of the opinion that increasing the quality of our tags, and maintaining a certain quality is important (otherwise, why have tags at all?), and that doing so would be greatly aided by increasing the minimum rep required to create them in the first place. I can go through and remove the hundreds of tags that ought not exist and have only a question or two in them, but that does little good if hordes of low-rep users come in behind me and decide to tag their questions with what are essentially 3 or 4 important words from their particular problem. Off the top of my head, I think I'd place the tag creation privilege at or near the tag synonym creation privilege, which is 2,500.


<Sidebar>

On the off chance the purpose of tags is not clear to anyone, tags exist to categorize and organize. We currently have over 5,000 tags. Over 1,000 of those have only a handful of questions (5 or less), and the median number questions per tag is approximately 16. To me, this indicates that at least a large number of our tags are of poor quality.

IT is a huge field, but I find it hard to defend a categorization system which divides it into over 5,000 categories... and all the more when thousands of those thousands of categories only apply to a handful of questions out of the >180,000 questions we have on the site.

The only explanation I have is that we have a very large number of tags that should not exist, because they are not actually IT-related categories. (To borrow from my bad tags thread, "fixed" is not a category related to IT. "Width" is not a category related to IT. Nor is "rotating," "outbox," or "social." Not even if the question is about fixing the width of your rotating inbox while remaining social with your co-workers.

The problem with having large numbers of poor quality tags is that they defeat the benefits of categorization that tags exist to create in the first place. Searching through 5,000 categories is harder and more time-consuming than searching through 4,000, and if those "extra" 1,000 tags don't have any value, then you've just inflicted increased cost and effort on everyone for no reason.

</Sidebar>


Having said all that, what does anyone else have to say on the subject? I'm thinking I should request of the Stack Exchange CMs that the tag creation reputation threshold be raised to 2,500 for Server Fault.

Yes? No? Too high? Too low? Don't Care?

3
  • I agree with this (but also with Iains reduced number), but do we have any chance to get this done by the CMs? I might be wrong and stopped caring at some point, but I always had the impression that privilege thresholds on all network sites where modeled after the activity pattern of Stack Overflow and carved in stone afterwards...
    – Sven
    Feb 13, 2015 at 11:38
  • 1
    @Sven if you look on mSE, you can find a couple threads were the SO folks suggested (and got) the rep threshold bumped for SO and mSO. With SO being a special snowflake and all, that no guarantee they'll do the same for us, but it does show that it's at least technically possible for them to do so. Feb 13, 2015 at 14:33
  • 1
    Just don't raise it beyond my rep :)
    – MDMoore313
    Feb 23, 2015 at 1:32

3 Answers 3

23

Yes increase the reputation requirement to create tags. Most people with only 300 rep probably don't quite get how tags are supposed to work. People tend to forget or not know that spaces separate tags and where a - should be used or that they are sorted by popularity. For example the and you cite should probably be and they could end up as , . There are plenty of other examples of this.

I think 2500 is probably too high. Rep is hard to come by on SF perhaps 1000 or 1500 would be better.

5
  • 4
    Taking myself as an example, I think 1000 is a high enough threshold. I'd find it odd/unfortunate if I was unable to create tags...
    – Reaces
    Feb 13, 2015 at 10:57
  • 7
    +1 for 1000 to 1500. There's a tradeoff here between eliminating worthless tags and granting access to a useful organization tool. I'm of the opinion we'd significantly reduce lousy tags at 1000 rep without putting new tags out of reach of the majority of contributors. Feb 13, 2015 at 13:51
  • 5
    Another +1 for 1000-1500 here. I think that users with that rep must know what a tag is and how it should be used.
    – Frederik
    Feb 14, 2015 at 12:36
  • 4
    I like 1000, too. Feb 14, 2015 at 14:50
  • 20 percent is not a bad ratio for a taxonomy without rules. I suspect that further limiting the creation of tags will likely mean untagged or mistagged items. It's probably easy to fix an item tagged "fixed", "width" as fixed-width, not so much when it's tagged "unix"
    – Jim B
    Feb 19, 2015 at 4:13
10

I baited a friendly and unsuspecting Community Manager just now.

The request to increase rep minimum (to 1000, or failing that, 1500) for the tag creation privilege has been made and received.

2
  • 4
    I'm imagining an unsuspecting community manager at a watering hole in a nature documentary. Run little gazelle, run! Feb 17, 2015 at 3:42
  • 1
    I never thought of the Noob as a gazelle but those CMs are really something ;)
    – user9517
    Feb 17, 2015 at 15:37
6

So... Normally, I would collect some data on how many tags are created by users at various rep levels, and how they were used. This is a pretty decent way to gauge whether anything of value would be lost if the privilege were bumped up.

I started to do that here. The results looked good - barely any tags created by low-rep users, a slam-dunk . Then I ran the numbers on tags created by high-rep users, and things got weird: they weren't creating very many tags either; in fact, I counted only 87 tags created here during the last year by anyone. That seemed... Odd. So I went to do the last check: how many tags were destroyed during this time-period.

I found that only a hundred or so had been destroyed by the system due to low/no use. However... In the past couple weeks, over 1400 tags have been merged, mostly into this big ball of mud.

That... Pretty much makes doing any further analysis right now worthless. It'd be really nice if we tracked information on deleted tags, such as who created them and so on... But we don't.

All I know is that of the extant tags created during the past year, there's not a very significant split between low-rep and high-rep users; the only new tag to garner more than a few questions did so via merges and will be destroyed itself - if new tags must be created in the process, that will only skew this analysis further.

Marking this - please get rid of this tag as quickly as possible, and then remind me to check again in another few months to see who is actually creating useful tags here.

Update

Over a month later, that tag was still around. So I've removed it with extreme prejudice. There is some cleanup to do in a tag that cannot be added to new questions or ignored when editing old.

I will revisit the tag-creation privilege in approximately six months, by which time we should have sufficient evidence of need.

5
  • 2
    I've sliced it down from 10.6k to... the 9.8k it's at now in less than a week, so it'll be gone or reduced to a smaller collection of no-valid-tag, but historically-significant questions in... well, a few months or so, given current rates. Feb 25, 2015 at 3:28
  • Would it be possible to look at older statistics? What I mean is, because of the fairly recent start of the off-topic tag, a database backup from 2 months ago would allow for you to get valid statistics without the skewed / tainted results.
    – Reaces
    Feb 25, 2015 at 8:52
  • 1
    Even if I did that, @Reaces, it still wouldn't give an accurate picture. Although he did it in a profoundly odd way, Hopeless was performing some much-needed cleanup - a good number of the tags that existed two weeks ago probably shouldn't have. The problem is, right now we don't know which ones were needed and which ones weren't - once we've sorted through the 8K or so questions that currently have NO valid tags and given them valid tags (creating new ones as needed) then we'll have a better picture (unfortunately, one not connected to the data that previously existed).
    – Shog9
    Feb 25, 2015 at 16:52
  • 1
    @Shog9 I don't think I fully understand. As the re-tagging and tag-cleanup efforts in part caused this request, wouldn't it be best to review the situation before the clean-up? As HopelessN00b indicated that a possible reason for the bad tags he's cleaning up is the lower rep threshold required to create them. It would be kind of ironic that the cleanup efforts end up with a picture of only good tags from low rep users. While data from before december would clearly show the hundreds of garbage tags with only a handful of questions that he cleaned up since.
    – Reaces
    Feb 26, 2015 at 8:47
  • 2
    I suppose it's too late now, and perhaps this was the point, but I'd been using that tag as a group of questions that needed review and action, and at least one normal user had been using it to mine flags (VTC), since it was primarily admin panel questions, with a significant minority of other atrocious, unsalvagable tags. Oh well. Apr 10, 2015 at 13:50

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .