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.