While I expect most women buy in advance, I expect they also don't buy enough and so are out there mid-period at least some of the time. Their cycles also affect their buying patterns, so even women who are stocking up are more likely to buy differently in different parts of the cycle.

The real question is what does the manager do with the data. Across a city on any given day there are about the same number of women in every day of their cycle so it isn't like they are marking up pads on the 25th-28th day of the month to get women who didn't stock up. As such I don't think this data is useful, what is useful is when they discover a women missed her cycle and thus needs to get ads for pregnancy wear in a couple months. Since that is their need for data, the fact that it is noisy and not very accurate is still close enough.

That said, they probably are more accurate than you would expect despite all the noise. Not 100% accurate, but being greater than 50% accurate is a lot better than chance and should be obtainable.