Only when you have a known population that you can survey. You can survey a city. You can survey the attendees of a movie, play, sports game, etc.
You cannot survey an unknown population. When your product is a set of three books, and n number of people in a group need only one set, you cannot tell how many people are there, what their sex is, or how often they play.
The numbers presented here are about as accurate as randomly generated values. They have no way of knowing how many people play, no way of contacting them, they're blind.
You are engaging in a false dichotomy. In fact, a city isn't a simple known population, and this one isn't an entirely unknown population. In fact there are ways to determine population sizes if you want to know that, for instance
capture-recapture methods, which don't actually require ear tags or leg bands to make work. The reality is that early questions in the survey are used to filter out people who are ineligible and shouldn't be counted. I'm sure there are a few people who slip in or get excluded, but even then there are methods to correct for this reasonably well.
The closest thing they could do is cold call people, or survey some game shops. The most likely thing they could do is survey game shops in their HQ region, which given the culture in their area, is going to give them *extremely* bad information. I don't think anyone would try to assert that Seattle is equivalent to anything other than a few cities in CA, and very different from everywhere else.
No survey will be perfect, but given some triangulating information it's quite possible to reality check it. A responsible survey analyst will do these things. There are also ways to check the response rates against hard(er) numbers, such as sales figures, return figures, numbers from the book trade vs game shop distributors vs online, etc. These are things that publishers check all the time. I know for fact that WotC can because they are distributed by Random House.
Now, there are ways they could be very, very wrong, typically when the sampling is flawed. For instance, their Unearthed Arcana weekly surveys were clearly quite flawed due to the fact that they were totally volunteer with no attempt at outreach beyond what was posted on their web page. So if WotC is trying to determine things like the numbers they report from surveys like that, it would be really bad.
As I said in another post, I don't know who WotC hired to do this work and it's possible they didn't do it properly. Certainly you can't count on WotC to be anything but self-flattering with their announcements. However, survey companies like Qualtrics and YouGov really do know what they're doing and the principles are pretty well understood. I took an
entire doctoral class in grad school on the topic despite survey research not being my area. People get their PhDs on it and spend their entire career working in the area. These are precisely the people that Qualtrics and YouGov hire.