You run a 28 day autocorrelation and see what pops out. Start with the obvious purchases then artificially synchronize the purchase history of a group of women to be on the same phase and run another autocorrelation on their aggregate purchases to get the non-obvious signal.

I would imagine women who are buying menstrual products are very likely either about to begin their period or are in the midst of it. Also, things like midol are probably strong indicators.

Hmm, I'm not sure–all the women I know buy massive packs from Costco because it's something that they might as well buy in bulk and save money on. I didn't think of medicine, though–you're probably right about that.

Painkillers and chocolate purchases should be slightly correlated.