You should be good with something like that (suppose df is your data):
df.agr=aggregate(Product~Customer,data=df, FUN=function(x) length(unique(x)))
df=cbind(df, Count=apply(df, MARGIN=1, FUN=function(r) df.agr$Product[match(r[1],df.agr$Customer)]))
It won't be blazing fast, but definitely faster than for.
