If they had over 10,000 Receipts showing one or the other winning, This is actually a pretty large poll and I wouldn't expect it to be over 5% error, maybe 10% from that poll. However I'm not sure what there actual sample size was, obvious the bigger the better.

Depends on what you mean by respectable. For the Wii U, would it be respectable to just show good YOY growth even if they're still not in the same ballpark with the other 2?

Sample size is adequate, but there may be selection bias and other things that make the data inaccurate.

Yes for the Wii U it would, since I take into consideration the lowered expectations of a system that must rely on mostly first party support in order to sell.

real real real plot twist: Ouya isnt real. its a troll the internet masters are trolling us with.

You could have 100,000 receipts, but that doesn't necessarily mean anything if the sample isn't random.

For example, when I used to work retail, I worked in a community with a largely Russian population.  Even when the Wii was dominating and the PS3 was sucking wind, we always had more PS3 sales.  So, if you took receipts from our store, it would give you a really skewed view.  There is no reason to expect it to be accurate to 5% unless you have some knowledge of their sampling method.  Having worked on Black Friday, I can't imagine the sampling method was anything more complex then checking random receipts.

This app would cover the whole country wouldn't it? Anyone can download it. So I'm sure it's pretty random.

It's no more accurate than an internet poll, and internet polls can be notoriously innacurate with way more than a 10% error. A 5% error is a not too bad error from a well designed poll, and this is not a well designed poll. Apparently last year they had 1500 console hardware receipts, from something like 120K receipts over all, and that would have included handhelds. If they received 185K total receipts this time that's a ~50% increase in number of receipts, so the console hardware sample size might be 2200-2500. Also the samples are exclusively limited to Target and Walmart. So while they may be a reasonable approximation of console sales for those two retailers it would be wrong to extraoplate that to be a credible reflection of all retailers.  

the only thing I know is: Wii U >>>>>>>>>>>>>>>>> PS3

That research firm did note that most of the Xbox One buyers were families and most of the PS4 buyers were not. What if families are more likely to use this app than young adults? They also only take from Target and Walmart. Again maybe families are more likely to shop there? There are many selection biases. I still think that the Xbox won (by a decent margin too), but unless the data collected is randomized you can't really use them.