Ran a few test queries on en.wiki (all of these are constrained to accounts that are at least 1 year old and aren't bots):
Article namespace (>=100 edits in last 30 days): 2819 users
File namespace (>=100 edits in last 30 days): 28 users
Draft namespace (>=100 edits in last 30 days): 35 users
Project namespace (>=5 edits in last 30 days): 2013 users
MediaWiki namespace (>=5 edits in last 30 days): 13 users
Template namespace (>=5 edits in last 30 days): 1496 users
Category namespace (>=5 edits in last 30 days): 523 users
Portal namespace (>=5 edits in last 30 days): 95 users

Same queries on Arabic Wikipedia:
Article: 90
File: 3
Draft: N/A
Project: 74
MediaWiki: 2
Template: 70
Category: 51
Portal: 21

Perhaps we should cut the thresholds in half for smaller wikis. We still aren't going to get a comparable sample, but hopefully we'll at least get enough people to have usable data on those wikis.

We need to get a hard number for target pool sizes on each project.

Here's one of the SQL queries I used to generate the numbers above:

select rc_user, rc_user_text, user_id, user_registration, count(*) as cnt
from arwiki_p.recentchanges, arwiki_p.user
where user_id = rc_user and
rc_namespace = 14 and
rc_bot = 0 and
user_registration < DATE_SUB(NOW(),INTERVAL 1 YEAR)
group by rc_user
having count(*) > 4
order by cnt desc
limit 10000;

Waiting to hear back from @egalvezwmf on our target sample sizes.

No itwiki? 😕

Here are my suggestions for the 5 "top" projects

Your sorting is broken. Top X are defined by usage and as seen on https://www.wikipedia.org/ the top 5 are currently en, es, de, ja, ru.

When adding Commons to the mix, that makes en, es, de, ja, Commons.

In T107829#1508020, @kaldari wrote:

Ran a few test queries on en.wiki (all of these are constrained to accounts that are at least 1 year old and aren't bots):

Your selection of namespaces appears to be arbitrary, what is the rationale? Please use WikiStats to comply with standards, i.e. content namespaces vs. non-content namespaces. No need to make queries.
https://stats.wikimedia.org/EN/TablesWikipediaEN.htm#editor_activity_levels

@Nemo_bis: I was sorting by article number since Wikipedia usage overlaps more with internet usage (the criteria for the "strategic" projects). Either way, it seems to result in the same 10 wikis. For example, if we take the 5 "top" wikis (by usage) and the 5 top internet languages and combine them, we get:
English
Spanish
German
Japanese
Russian
Chinese
Arabic
Portuguese

If we then add Commons, that gives us 9 projects.

The next project by top usage or article count or internet language is French, which gives us 10.

Your selection of namespaces appears to be arbitrary, what is the rationale? Please use WikiStats to comply with standards, i.e. content namespaces vs. non-content namespaces. No need to make queries.

Thanks, I didn't think to look at stats.wikimedia.org. That's a lot easier :)

@kaldari, the sample size will come later; hopefully by tomorrow. We still need to translate. Thanks!

also: http://wikistats.wmflabs.org/display.php?t=wp which lets you sort by good/total/edits/users by clicking the little arrows in the column header

Looks like we're shooting to invite ~100 users from each of the 10 projects.

Since we will probably be conducting this poll more than once and we want to avoid asking the same editors to take it multiple times, we will need to create a pool of eligible editors and then select randomly from that. My suggestion for the selection process would be:

For Wikipedias:
Select the top 100 main namespace editors
Select the top 300 non-content, non-talk editors
Combine the two lists
Remove duplicates
Randomly select 100 to invite

For Commons:
Select the top 100 file namespace editors
Select the top 300 non-file, non-talk editors
Combine the two lists
Remove duplicates
Randomly select 100 to invite

Wikipedia usage overlaps more with internet usage (the criteria for the "strategic" projects).

So why don't you just take the top 9 Wikipedias + Commons? An internal source for internet usage sounds better than the unverified external source.

Besides, the English Wiktionary has as many pageviews as the 12th most visited Wikipedia, let's not give for granted that Wikipedia is the only Wikimedia project with sizable userbase.

That was basically my original idea, but the Foundation wants us to look at "strategic" wikis as well, i.e. wikis that have strong potential for growth. In practical terms, that means including Chinese Wikipedia and Arabic Wikipedia instead of Italian Wikipedia and Polish Wikipedia. I actually think this is a good idea as those two wikis are likely to have more unique needs (such as dealing with an influx of new editors and content, as well as unique i18n issues). They are also wikis that the Foundation doesn't really know much about as we don't get as much input from them as we do from the European-language wikis.

The selection of projects here is related to the organization of the promotion of the survey by Community-Tech, right? I mean, if someone else wants to volunteer their work promoting this survey in other projects, are they free to do so or is there any limitation?

@Qgil: There are actually 10 separate surveys. Each one is customized for the specific language project (for example, using the local name for WikiProjects). I think it would cause confusion to promote the survey on other projects.

In T107829#1505277, @kaldari wrote:

The best way to generate this list would probably be to query the recentchanges tables for users who edit certain namespaces the most. For example, users who are at least 1 year old and meet any of the following thresholds:

• More than 100 edits to the Article, File, or Drafts namespaces in the past month
• More than 5 edits to the Template, Module, Project, Category, MediaWiki, or Portal namespaces in the past month

These numbers should be tweaked to get a good balance of content curators and technical editors. They may also need to be significantly lower for the strategic languages in order to get a large pool of editors to survey.

@kaldari, can you confirm if this criteria still holds? And the wikis we are targeting are:

1. English
2. Spanish
3. German
4. Japanese
5. Russian
6. Chinese
7. Arabic
8. Portuguese
9. Commons
10. Italian

Is that correct?

French, not Italian.

• More than 100 edits to the Article, File, or Drafts namespaces in the past month
• More than 5 edits to the Template, Module, Project, Category, MediaWiki, or Portal namespaces in the past month

I'm wondering how the numbers for each of those 2 bullet points compare. I don't want the lists to be dominated by article writers rather than technical contributors. Would it be possible to do a test query for each of those 2 on en.wiki to see what the numbers are?

Actually, I have a better idea. What do you think about querying for:

• The top 100 contributors to the content namespaces (Article, File, or Drafts) in the past month
• The top 100 contributors to the technical namespaces (Template, Module, or MediaWiki ) in the past month
• The top 100 contributors to the curation namespaces (Project, WikiProject, Category, or Portal) in the past month

Combine the 3 lists, remove duplicates, and randomly select 100 to invite.

That should work for wikis of all sizes (unlike my previous suggestion).

Sounds good to me.

Nitty-gritty time! (The namespaces listed below are meant to be representative, not definitive, i.e. not every namespace is include that possibly could. This should be good enough to give us a representative sample of users across different areas of the projects.)

• English Wikipedia
  • Content: 0, 6, 118
  • Technical: 8, 10, 828
  • Curation: 4, 14, 100
• German Wikipedia
  • Content: 0, 6
  • Technical: 8, 10, 828
  • Curation: 4, 14, 100
• French Wikipedia
  • Content: 0, 6
  • Technical: 8, 10, 828
  • Curation: 4, 14, 100, 102
• Russian Wikipedia
  • Content: 0, 6, 102
  • Technical: 8, 10, 828
  • Curation: 4, 14, 100, 104
• Commons
  • Content: 0, 6
  • Technical: 8, 10, 828
  • Curation: 4, 14
• Chinese Wikipedia
  • Content: 0, 6, 118
  • Technical: 8, 10, 828
  • Curation: 4, 14, 100
• Spanish Wikipedia
  • Content: 0, 6
  • Technical: 8, 10, 828
  • Curation: 4, 14, 100, 102
• Arabic Wikipedia
  • Content: 0, 6
  • Technical: 8, 10, 828
  • Curation: 4, 14, 100
• Portuguese Wikipedia
  • Content: 0, 6
  • Technical: 8, 10, 828
  • Curation: 4, 14, 100
• Japanese Wikipedia
  • Content: 0, 6
  • Technical: 8, 10, 828
  • Curation: 4, 14, 100, 102

You could actually run the same 3 queries against each of the wikis except for Russian (as that's the only wiki that doesn't have namespace alignment with the others):

• Content: 0, 6, 118
• Technical: 8, 10, 828
• Curation: 4, 14, 100, 102

The technical namespaces are the same for all of them, even Russian.

Here's the output of my queries:

Let me know if I should format it better (with User: prefix?).

Here's the code used to generate it: https://github.com/Niharika29/database-reports/blob/master/survey.py (I need to move it into a new repo, but for now)

@NiharikaKohli: We'll actually need this as 10 separate lists, 1 for each wiki. This is because there will be a separate poll for each of the wikis that is customized for that language and project. If someone gets invited to take more than 1 of the polls, that's OK.

@kaldari, these are actually ten lists, just combined into one paste. The first line indicates the wiki followed by hundred users of the wiki for all the ten wikis.

Ah, got it!

Maybe I'm late for this, but I think that for reaching people with "technical opinions", you should add the operators of the most active bots in those projects; programming is also an important technical work which leaves less time for editing, even in "tech namespaces". Yes, I'm an example of that, and I don't appear in P2064; and yes, it won't be so easy to extract that list of operators. If you consider adding them by hand, I can give you a list for eswiki gladly.

@-jem-: I would really love to have bot writers included in the survey. Unfortunately, I don't know of any way to consistently generate a list of active bot writers for each project. (We have to be able to periodically generate a list with the exact same algorithm each time so that the data is comparable from year to year.) Unfortunately, it isn't possible to query for bot owners in the database, which is a big limitation. If you have any ideas, let me know.

@kaldari, yes, I already stated that there is no easy way. It is possible to join the recentchanges and user_groups tables to obtain the most active accounts with the bot flag (the unflagged bot accounts will usually be editing at a "human" rate and, as such, much less active), but to get the operators it would be needed to check the user page of the bot and search for a parameter of a bot-identifying-template, or something like that, very project-depending anyway. In eswiki we solved the problem by creating https://es.wikipedia.org/wiki/Template:Controlador, which returns the operator and also allows us to identify the unflagged bot accounts, but of course it must be updated by hand. For just 10 projects, as in this case, for sure it is faster to ask a trusty user from each one for a list of active bot operators, and that's why I offered for that; but, of course, I don't want to delay any schedule. I'll be glad if I can help and also if it isn't needed or there's no time for it.

It would be vastly easier to write on the bot accounts' talk page, which is usually either watched or redirected to the owner's. Someone might get a duplicate but that's not a huge issue.

For example, https://it.wikipedia.org/wiki/Template:RichiestaBot used at https://it.wikipedia.org/wiki/Wikipedia:Bot/Richieste supports "operatore" as parameter.

@Nemo_bis: I think your suggestion sounds the most practical. I'll create a separate task for it since we'll need to send them slightly different invitations anyway.

In T107829#1555203, @kaldari wrote:

Here's one of the SQL queries I used to generate the numbers above:

select rc_user, rc_user_text, user_id, user_registration, count(*) as cnt
from arwiki_p.recentchanges, arwiki_p.user
where user_id = rc_user and
rc_namespace = 14 and
rc_bot = 0 and
user_registration < DATE_SUB(NOW(),INTERVAL 1 YEAR)
group by rc_user
having count(*) > 4
order by cnt desc
limit 10000;

You are aware this includes wikidata actions?

@Bawolff: What do you mean by a "wikidata action"? Can you provide an example?

e.g. The things on https://en.wikipedia.org/w/index.php?namespace=14&tagfilter=&hidewikidata=&title=Special%3ARecentChanges that have a D beside them (rc_source = 'wb' )

I've updated the paste as per the new query: