We spend a fair amount of time talking about question migrations between sites. These conversations happen internally, publicly and semi-publicly in the network-wide moderator chat room. In the interest of reducing the time we collectively spend discussing it, let’s see if we can formulate a clear and useful migration strategy.
Can we build a wall between sites?
My first thought was to just block migrations altogether. We already do this when questions are older than 60 days, so there’s some precedent for it. Certainly, it will only have a minimal impact on Stack Overflow1:
Migrated To site Away from site Migration % Site -------- ------- -------------- ----------- ---- 1498 728 770 0.20 Stack Overflow
But there are a number of sites with a fair number of migrations that represent more than 3% of questions posted on the site in the last 90 days:
700 434 266 4.07 Super User 541 24 517 9.52 English Language and Usage 495 482 13 14.61 English Language Learners 410 157 253 4.55 Statistical Analysis 367 60 307 3.94 Server Fault 282 49 233 4.23 Electronics and Robotics 267 149 118 3.21 Unix and Linux 265 207 58 0.62 Mathematics 204 251 79 5.72 IT Security
I looked at the ELU => ELL connection in particular. Organic asking accounts for most questions on the Learner's site. However several hundred are imported quarterly from the general English site. And this occurs with minimal friction as far as I can tell. Despite my subjective impression, question migration generally works, so we ought not take it away.
When does it makes sense to migrate?
In discussions among the Community Managers, we've expressed two mutually exclusive2 ideas of when to migrate questions:
Don’t migrate with a lot of history since that tends to cause problems on the target site.
Don’t migrate questions without a lot of history since there's nothing to lose by asking people to reask elsewhere.
Looking at the questions migrated away from English Language and Usage (mostly to English Language Learners), it seems they are often not answered or much upvoted before migration:
Migrated away Avg score Avg answers Avg comments Avg age (days) ------------- --------- ----------- ------------ -------------- 517 0.388781 0.552742 3.299168 2.865096
On the destination site, they tend to be voted on and answered:
Migrated here Avg score Avg answers Avg comments Avg age (days) ------------- --------- ----------- ------------ -------------- 482 1.192946 1.445378 3.455525 2.322613
Obviously, we are preselecting for newish questions since we don’t allow migrations after 60 days. I’m also assuming this is healthy site interaction. Looking at recently migrated questions, it seems the pattern is to limit interaction on these questions to comments (which often point out the off-topicness). In particular, people seem to be avoiding voting on or answering questions likely to be migrated.
Proposed way to think about migrations
Most of the time when we talk about migrating questions, it’s in the context of rescuing content otherwise doomed to deletion. But we already have a ready solution to that problem: historical locks3. Now that old questions can only be migrated by employees, migration just isn’t a practical tool for preserving content. So I’d like to propose an alternate theory of migration:
Migrate questions when it saves the asker the effort to reask.
In other words, migration is a service we offer to people who misunderstand the complex structure of Stack Exchange sites. Rather than forcing them to create an account on a new site, copy and paste their question (including title and tags), and potentially get comments and answers from two different sites, we just move the whole thing to where it belongs.
This is the philosophy behind the guidance4 we wrote for moderators:
Please don't "horse trade" questions. Don't migrate crap and remember that destination sites can reject migrated questions by closing them. If you still think a question needs migrating, follow these guidelines:
If the question is on-topic for the site where it was asked, and it is answered, decline immediately. (If you feel like being generous, check if the flagger is also the answerer and migrate if this is the case.)
If the question is off-topic or unanswered and the flagger has a good bit of reputation on the target site, go ahead and migrate; they probably know what they're talking about.
If the question is off-topic but seems reasonably well-written and you understand it well enough to believe it belongs on the target site, migrate.
Don't ping another site's mods about potential migrations unless you're honestly interested in learning more about their site's scope. Be sure to read the site's help/on-topic page.
I'd love to get your feedback on this proposed migration philosophy. But there are a few things it's probably not constructive to focus on:
There are some UI concerns with migration right now. While changing the interface to match the goals of the feature is good idea, let's nail down the purpose of migration first.
If you've seen an example of a bad migration, it's only helpful to bring it up if it illustrates a principle not considered above. My subjective judgment based on years of anecdotal evidence is that migration isn't working. But looking at aggregate data, I know that's bad analysis.
So what do you think?
The numbers from public data don't match the numbers below because the public data doesn't retain post history information about deleted posts. The numbers are also from a few weeks ago when I wrote the draft of this question. But these are representative.
I sometimes get the feeling that users assume CMs have a unified view of how things should work on our sites. That's a carefully crafted illusion created by having intense arguments in private and then having the winning view written up for public consumption. Many times we've talked about recording our conversations to publish as a podcast or somesuch.
Unfortunately Meta sometimes makes the internet just a little bit worse. A Google search for "historical lock" turns up: What is a historical lock, and what is it used for?
If you are a moderator, please see
/help/mod-tlon your site. That article also includes information on the Teachers' Lounge, which is valuable resource you might have forgotten about.