Re: Still confused about the 90%
There are hotspots of activity, some people live farther out than others, some have different travel routines, some work from home while others go to an office, etc. I think this is more what I was trying to say, and my paint metaphor may have been misleading.
I understand what you're saying, and can see why it might seem like that, but the randomisation also means those factors are spread evenly.
This is why I keep comparing it to a coin toss, because, sure it could happen that every single person in the treatment group lives in the countryside and works from home, and everyone in the placebo group works in an ICU, but it's exceptionally unlikely, you calculate the probabilities by multiplying and at first it starts out plausible, 50:50, then 25:75, then 13:87, 06:94, even up to the first ten it's about a thousand to one for that kind of imbalance and it's not utterly implausible. But it's exponential, so by the time you've got over eighty in one group and under ten in the other it's really unlikely it can happen by chance.
(For purists, I've skated over the combinational aspect if you allow a few of the placebo group to be living in the country, but as we're at the tails of the distribution it doesn't make much difference.)