We want to understand what will happen when we remove the problematic beta opt-out migration.

• Only look at the small default wikis: group0 + dewiki, fawiki, and arwiki.
• Split analysis by wiki
• Only consider editors who made at least one edit in July 2020. We don't want to dilute the results by including inactive editors. (FWIW, "active editors" typically means 5 or more edits in a month, but we'll use a threshold of 1 here.)
• Find the current percentage of active editors who are configured to use TwoColConflict.
• Find the percentage of active editors who would have been configured to use TwoColConflict, if the migration were not in place.
• Find how many users saw TwoColConflict in this month.
• Estimate how many more users would have seen TwoColConflict without the migration in place

Results

https://gitlab.com/adamwight/conflict-query/-/blob/optout/reports/Opt-out%20modeling.ipynb

	wiki 			active_editors 	beta_only_optouts 	full_feature_optouts 	anons 	beta_only_optouts_pct 	full_feature_optouts_pct
130 	dewiki 			51632 		477 			2603 			32257 	0.9 			5.0
131 	fawiki 			16005 		48 			288 			10763 	0.3 			1.8
132 	arwiki 			15243 		65 			257 			9060 	0.4 			1.7
57 	mediawikiwiki 		1454 		33 			111 			696 	2.3 			7.6
71 	officewiki 		177 		2 			42 			0 	1.1 			23.7
92 	testwiki 		173 		7 			33 			35 	4.0 			19.1
93 	testwikidatawiki 	49 		1 			3 			20 	2.0 			6.1
91 	testcommonswiki 	10 		0 			1 			1 	0.0 			10.0
84 	simplewikibooks 	1 		0 			0 			0 	0.0 			0.0
112 	wikimania2009wiki 	1 		0 			0 			0 	0.0 			0.0

Only "beta opt-out" users are affected by the migration code. This group is currently seeing the legacy interface, and after we remove the migration will see the TwoCol interface.

Note that the number of people seeing the legacy interface is much higher than can be seen in this table. In August, 41% of conflicts on dewiki used the legacy workflow. We think this is a consequence of more active users being more likely to have run into the Beta Feature opt-out bug.