8

1

We're thinking about putting together a "leaderboard" on StackExchange.com to:

  1. Help people find new / interesting StackExchange sites that are active
  2. Give StackExchange admins something to compete over :)

We're planning on having 3 or 4 columns with a list of the top 5-10 sites (expandable or pageable to see more). We'll recalculate the columns nightly and update the lists.

The question is, what should the columns be?

Some stipulations:

  1. We don't want it to be the same top sites in every category, so it shouldn't be biased towards any single criteria

  2. We'd like to see a mixture of existing sites that are going strong, and up-and-coming sites

  3. It has to be something we can easily calculate, so it has to be based on information we're already tracking somewhere (but not necessarily aggregating)

  4. The categories should be based only on the last week of data (or possibly week-over-week)

  5. They should measure things that actually correspond to an active / growing site, and be things we want to encourage (and not encourage bad behavior or gaming)

I'll post some of our ideas as answers so people can vote / comment on them.

Obviously this won't be perfect, and not every site will be make it onto the first page. We'll add a couple things to help that: 1) we'll have an element that just shows some random sites, so everyone gets a little exposure, and 2) if it works, we could expand the Community side of the site to include some sort of searchable/browsable directory. If you have more ideas for this we'd love to hear them.

Note: this will be strictly opt-in (you'll check a box in your account settings), so if you don't want your site to appear it won't

Update:

The leaderboard is now live! You can still add your site under admin | content (updated daily).

flag

12 Answers

6

Reputation Earned per user: total reputation earned by all users divided by the number of non-anonymous users that visited in the last week, with some minimum number of users (say, 25)

link|flag
4 
I'd discount the swathes of users with 1 rep from this – Rich Seller Feb 23 at 19:34
1 
Potential risk: this can be "gamed", and therefore can have counter-productive effects. Scenario: SE sites start to inflate reps to appear in "weekly charts"... visitors are disappointed by contents of top-rated SE sites. – tucson Mar 1 at 11:05
6

Answered Questions: percentage of questions asked in the last week that have an accepted or upvoted answer (at least 25 questions)

link|flag
5

Traffic: unique visitors in the last week

This one will obviously favor big sites, but it's also the goal of (almost) every StackExchange and the simplest metric for finding an active community.

Note: the first iteration of this might just be total pageviews, since we don't track uniques right now

link|flag
4

How about % change in views week on week? (or month on month to smooth out fluctuations) as it gives an idea of how the site's reach is changing.

link|flag
I like this derivative or growth measurement. – Chris W. Rea Mar 1 at 18:52
3

New Users: number of users that posted their first question or answer in the last week

link|flag
Active users, as I have seen some sites @ the top of the listings(meta.stackexchange.com/questions/2750/…) but there numbers are suspicious with the rest of the big sites.. So the amount of activley contriubuting users, just a thought.. – debug Feb 23 at 20:08
@debug you could make this a separate answer since it's really a different metric – David Fullerton Feb 23 at 20:13
2

Growth (%), calculated as the arithmetic average of week-on-week growth for each:

  1. Unique visitors
  2. New users
  3. New questions / or perhaps 'Students'
  4. New answers / or perhaps 'Teachers'

Thus ranks up-and-coming sites, but older and more established communities will likely not feature.

Exclude any site with less than 30/40/50 users to avoid the statistical noise of boot-strapping and sock puppets.

link|flag
2

I've said this before, but the hardest metric to fake is Google PageRank. You can also get an interesting metric from Hubspot's website grader. You should be proud of yourselves because StackExchange sites actually measure up quite well there.

link|flag
1

I would be in favor of a full directory of all opt-in SE sites, with a classification by category.

I personally do not see the value of the proposed system of columns.

Let me explain: I think you can have a site with just 1 active user, almost no pageview, and yet it can be very helpful... if ONLY when you post a question you get a helpful answer within a reasonable timeframe for the question-asker. For me that's a sufficient criteria. Any other criteria is, IMHO, noise.

I would let users ask the question where they wish. Users will come back if they get good answers.

(Thank you David for starting the discussion; I would love to have a directory to link to from my SE site)

link|flag
I think you're looking for this: stackexchange.com/directory/list – Michael Pryor Mar 4 at 16:48
Hi Michael, thanks for the link. Though it's not classified by category... What about users who are looking for a place to ask a question about something? How do they find their way to the proper SE site? – tucson Mar 4 at 18:47
@tucson, what categories would you choose? – Rich Seller Mar 6 at 17:59
@Rich I would pick the same categories we have in the list here in meta: meta.stackexchange.com/questions/4/… (Business, Education, Health, etc) – tucson Mar 7 at 13:19
1

To me, probably a metric most close to the Stack Exchange philosophy is a measure of user engagement. I've no idea what it would look like, but probably something like [(total number of votes per week) + (total number of questions and answers per week)] / total number of views. This would give you some idea of how engaged people are with the site and how many are just passers by.

To take it up another step, you ideally want engagement to gradually increase with the amount of content of the site, rather than just staying flat as content increases. You might therefore have something about the rate of change of engagement. Sites with high rates of change of engagement with time would be the one's to watch for the future. Decreasing rate of change of engagement would be a reasonable warning sign that the majority of the content on the site is produced by a few people and so the content production is unlikely to be sustainable.

link|flag
3 
I like the general idea of an engagement metric, but views as a denominator seems odd. A decreasing engagement by this measure could well indicate that the site has been viewed more, how is that a bad thing? – Rich Seller Mar 1 at 14:15
I would say that increasing views is not necesserily a bad thing, but you can have a situation with increasing views and decreasing angangement which is one of those mixed situations where both good things and bad things are happening at the same time. – Ian Turner Mar 1 at 14:48
1

Traffic: Total Pageviews (preferably by month)

Yes, the simplest metric that can be, I think it can actually be useful on the Directory.

It's the most direct way to see how well a SE site is performing.

Right now, the best metric to see how well a SE is doing is Users/Day, but I suspect there are older SE sites that get less users/day than us but actually have more traffic (hits from Google, more referral sites).

I saw David mentioned this metric in an answer (Traffic: unique visitors in the last week), but it didn't quite made it.

link|flag
0

Active users, as I have seen some sites @ the top of the listings, but there numbers are suspicious with the rest of the big sites.. So the amount of actively contributing users, just a thought.

link|flag
-2

Is the leaderboard up yet? If so, where?

I'd like to see how my SE site, PaleoHacks.com is doing relative to other SE sites.

link|flag
2 
really? did you not see the banner on the homepage? stackexchange.com/directory – debug Mar 4 at 16:50

Your Answer

Not the answer you're looking for? Browse other questions tagged or ask your own question.