Comparison between numbers indicating usage of fixed top bar vs dark mode looks somewhat inaccurate, I would suggest to at least "normalise" these numbers by some sort of site activity - by amount of views or by days visited or by reputation of users etc.

To start with, dark mode was announced much wider - respective announcement had featured tag allowing it to get 30K+ views, and two dedicated articles at SE blog: here and here, along with posts at multiple resources outside of Stack Exchange: zdnet, lifehacker, turnoffthelights just to name a few.

Compared to that, topbar announcement has never been featured and managed to gain mere 3K views, likely from regular meta visitors (0.015% does ring a bell?) - so it has got much harder start from the very start.

Another thing to take into account - which is maybe even more important - is that dark more is much more in-your-face and because of that much much easier to spread by word of mouth.

I can easily imagine someone looking over the shoulder of their colleague / classmate asking "Hey what's that, your Stack Overflow is dark, how can I make it look like that in my browser?" Chances of topbar tweak spreading like that are practically nonexistent in comparison because it is just so much harder to notice.

Suming up, these several thousands users aware of topbar switch more likely represent most active and motivated part of site audience ("core group" if you wish). Indiscriminately comparing them to hundreds thousands random passers-by feels kind of... slippery.

I think it would be safer to base your estimates on some kind of "weighted" metrics that would somehow account for site activity of involved users.

_{(Note to address particular clarification in comments - this is not specifically about dark mode and initial example with IE users was concerning for the same reason: simply counting amount of users while ignoring possible differences in their site activity is not a reliable way to estimate impact of the changes.)}

Comparison between numbers indicating usage of fixed top bar vs dark mode looks somewhat inaccurate, I would suggest to at least "normalise" these numbers by some sort of site activity - by amount of views or by days visited or by reputation of users etc.

To start with, dark mode was announced much wider - respective announcement had featured tag allowing it to get 30K+ views, and two dedicated articles at SE blog: here and here, a dedicated hat in the 2020 winter bash, along with posts at multiple resources outside of Stack Exchange: zdnet, lifehacker, turnoffthelights just to name a few.

Compared to that, topbar announcement has never been featured and managed to gain mere 3K views, likely from regular meta visitors (0.015% does ring a bell?) - so it has got much harder start from the very start.

Another thing to take into account - which is maybe even more important - is that dark more is much more in-your-face and because of that much much easier to spread by word of mouth.

I can easily imagine someone looking over the shoulder of their colleague / classmate asking "Hey what's that, your Stack Overflow is dark, how can I make it look like that in my browser?" Chances of topbar tweak spreading like that are practically nonexistent in comparison because it is just so much harder to notice.

Suming up, these several thousands users aware of topbar switch more likely represent most active and motivated part of site audience ("core group" if you wish). Indiscriminately comparing them to hundreds thousands random passers-by feels kind of... slippery.

I think it would be safer to base your estimates on some kind of "weighted" metrics that would somehow account for site activity of involved users.

_{(Note to address particular clarification in comments - this is not specifically about dark mode and initial example with IE users was concerning for the same reason: simply counting amount of users while ignoring possible differences in their site activity is not a reliable way to estimate impact of the changes.)}

Comparison between numbers indicating usage of fixed top bar vs dark mode looks somewhat inaccurate, I would suggest to at least "normalise" these numbers by some sort of site activity - by amount of views or by days visited or by reputation of users etc.

To start with, dark mode was announced much wider - respective announcement had featured tag allowing it to get 30K+ views, and two dedicated articles at SE blog: here and here, along with posts at multiple resources outside of Stack Exchange: zdnet, lifehacker, turnoffthelights just to name a few.

Compared to that, topbar announcement has never been featured and managed to gain mere 3K views, likely from regular meta visitors (0.015% does ring a bell?) - so it has got much harder start from the very start.

Another thing to take into account - which is maybe even more important - is that dark more is much more in-your-face and because of that much much easier to spread by word of mouth.

I can easily imagine someone looking over the shoulder of their colleague / classmate asking "Hey what's that, your Stack Overflow is dark, how can I make it look like that in my browser?" Chances of topbar tweak spreading like that are practically nonexistent in comparison because it is just so much harder to notice.

Suming up, these several thousands users aware of topbar switch more likely represent most active and motivated part of site audience ("core group" if you wish). Straightforward comparing them to hundreds thousands random passers-by feels kind of... slippery.

I think it would be safer to base your estimates on some kind of "weighted" metrics that would somehow account for site activity of involved users.

Comparison between numbers indicating usage of fixed top bar vs dark mode looks somewhat inaccurate, I would suggest to at least "normalise" these numbers by some sort of site activity - by amount of views or by days visited or by reputation of users etc.

To start with, dark mode was announced much wider - respective announcement had featured tag allowing it to get 30K+ views, and two dedicated articles at SE blog: here and here, along with posts at multiple resources outside of Stack Exchange: zdnet, lifehacker, turnoffthelights just to name a few.

Compared to that, topbar announcement has never been featured and managed to gain mere 3K views, likely from regular meta visitors (0.015% does ring a bell?) - so it has got much harder start from the very start.

Another thing to take into account - which is maybe even more important - is that dark more is much more in-your-face and because of that much much easier to spread by word of mouth.

I can easily imagine someone looking over the shoulder of their colleague / classmate asking "Hey what's that, your Stack Overflow is dark, how can I make it look like that in my browser?" Chances of topbar tweak spreading like that are practically nonexistent in comparison because it is just so much harder to notice.

Suming up, these several thousands users aware of topbar switch more likely represent most active and motivated part of site audience ("core group" if you wish). Indiscriminately comparing them to hundreds thousands random passers-by feels kind of... slippery.

I think it would be safer to base your estimates on some kind of "weighted" metrics that would somehow account for site activity of involved users.

_{(Note to address particular clarification in comments - this is not specifically about dark mode and initial example with IE users was concerning for the same reason: simply counting amount of users while ignoring possible differences in their site activity is not a reliable way to estimate impact of the changes.)}