Data source: June, 2011 (23 SE sites, not including Meta sites)
Question source: What can we learn about migrations?
Where do migrations originate, and where do they end up?
Sources (min 100) Targets (min 250)
Stack Overflow 24082 SuperUser 12856
ServerFault 3747 Programmers 6681
SuperUser 3660 ServerFault 6678
Programmers 618 Stack Overflow 2286
Webmasters 350 Webmasters 1480
Web Applications 260 Web Applications 1145
Unix and Linux 488
Apple 401
Android 327
WordPress 276
- The above results include deleted questions
- People who have said that SuperUser is a dumping ground for miscellaneous computer-related questions certainly have a case. (Have you hugged a SuperUser mod today?) Stack Overflow accounted for about 9,500 of the migrations, with the bulk of the remainder from ServerFault (this isn't surprising, but I thought it would be useful to include).
- While the data analysis did not specially consider the initial influx of questions to new SE sites, the most active sites in terms of migrations (say, Webmasters) also have a significant outflux of migrations as well, so IMO, while this does skew the stats a bit, it probably doesn't skew them as much as if those stats were run immediately after the initial influx. This would affect sites with lower absolute numbers, but those carry less weight anyway.
How well does a community know other communities? (Good outbound migrations)
Site Total Migrations Good Migration % Game Development 16 100.0 Theoretical CS 27 100.0 Unix and Linux 44 100.0 Programmers 618 98.9 Android 83 98.8 Ask Ubuntu 53 98.1 SuperUser 3660 97.2 Web Applications 260 96.5 Apple 56 96.4 ServerFault 3747 96.3 Webmasters 350 95.7 Statistical Analysis 71 94.4 Mathematics 52 94.2 Gaming 30 93.3 English 29 93.1 Electrical Engineering 41 92.7 StackOverflow 24082 90.3 Photography 33 87.9 WordPress 49 87.8
- Results from sites with fewer than 10 migrations-out were discarded
- A good migration is when a migrated question remains open on the target site, or is closed as a duplicate. In other words, "was the question a good fit for the target site?"
- Network-wide, about 9 out of 10 migrated questions end up in the right spot.
- Stack Overflow specifically had one of the lowest percentages and the highest volume -- increasing migration quorum was implemented as a result, so while that change is not reflected in these numbers, I suspect the good migration percentage from Stack Overflow has improved since then. Even so, 90% success is pretty astounding considering the number of people involved.
How well do other communities know our community? (Good inbound migrations)
Site Total Migrations Good Migration % English 26 100.0 Game Development 102 100.0 GIS 27 100.0 Unix and Linux 488 99.6 Apple 401 99.3 TeX - LaTeX 235 99.2 StackOverflow 2286 98.6 WordPress 276 98.6 Mathematics 95 96.8 Webmasters 1480 96.6 Statistical Analysis 83 96.4 ServerFault 6678 95.8 Web Applications 1145 94.8 Gaming 59 93.2 SuperUser 12856 92.8 Android 327 89.9 Programmers 6681 81.6 Electrical Engineering 48 79.2 Theoretical CS 14 78.6
- Results from sites with fewer than 10 migrations-in were discarded
- I think the staggering drop-off below the ~90% threshold is due to ambiguity of a site's accepted range of questions. It took quite a long time even for me to figure out what is acceptable on Programmers. Unfortunately, there isn't a lot of other data to support that conclusion, but I know that the "Theoretical xyz" sites in particular had come under fire for this. (See this, unfortunately.)
- Looking at the list of sites by name, I think we can conclude that good migration targets have a simple, clearly-defined range of acceptable questions. The less ambiguity, the better. This, of course, does not apply only to migrations, but probably to the site's overall success in general.
If you'd like to have a look at all statistics I collected for this question, including the full source/target migration matrix with close reason breakdown, you can download the Excel spreadsheet here.
