First, thank you for the info.
The main issue with how you presented the data is that a user's location does not indicate which layout they use.
For example, ISO users in Canada do not necessarily use a "Canadian" layout (they could be living abroad). Same goes for the US, obviously.
What I gather from this survey is that most ISO users (outside the US) are located in the UK, Germany, Sweden, the rest of the Scandinavian countries, and so on.
Besides, I rather go by recent group buy statistics than user location or layout. Because it's how many kits you think you will sell that dictates MOQ, not demographics directly.
Thank you for the feedback.
The country/layout breakdown should be taken at face value, because the dataset we have access to (as large as it is, with 1900 responses) didn't have a specific question for layout variants. Оf course, you're free to interpret the data however you like, and you're certainly correct that there isn't a 1:1 correspondence between a user's country of residence and the layout that they use. However, I don't think this makes any considerable impact on the result and conclusions of the analysis.
The important takeaway of the layout part of the analysis is that R3 \|, R4 <> seems to be the better option by a large margin, even when taking into account the potential discrepancy between location and layout (which almost certainly does exist). And, honestly, choosing to ignore the remaining data because of that discrepancy seems like intentionally choosing not to see the forest for the trees.
But let's give you the benefit of the doubt and consider a near worst-case scenario, where what you said is truer to a much larger degree than it is IRL. I'll primarily use the US as an example, since I think we can all agree that it's the most unexpected result, and there's a lot of expats living in the US (more than in any other country from the list). A lot of those expats are yellow-group-layout users to boot: there's a total of around 50–60 million Brits, Brazilians and others living in the US, Canada and Australia. Even if you take
all of the surveyed ISO users in the US,
remove them from the green pile, and
add them to the yellow pile together with users from Canada and Australia, it's still not nearly as large enough as the remaining users in the green group (it comes out to 56% vs 35% in favor of R3 \|, R4 <>). This is a near worst-case scenario, because we're assuming that none of the self-reported ISO users in the US use a US ISO layout,
and that all of them instead fall into the R3 #~, R4 \| category. Even then, R3 \|, R4 <> wins out by a considerable margin, which is pretty conclusive in my opinion.
The point is that the way it's been done in a lot of past group buys (usually R3 #~, R4 \| or R3 \|, R4 \|) is not perfect, and actual user data seems to back that up beyond any reasonable margin of error.
Just because something has been done a certain way in the past, doesn't mean that way is good. If that weren't the case, we would still be using Windows logos in all GMK sets. Change is good.
P.S. Unless the same set is run twice, each time with a different option for the ISO keys (but otherwise staying largely the same), looking at GB sales as an indicator for this doesn't mean much, since different sets/colorways enjoy different levels of popularity (and even the same set running at different times of the year, or at different points during the community's growth).