MY COMMENTARY:
Segmentation by measures of shopper visit recency, frequency of visits, and monetary expenditure, aka RFM analysis, is a tried and true activity of Customer Relationship Management. The premise is that customers who spend more, more often and more recently, are more valuable compared with others who shop less often and buy less. This works pretty well for B2B CRM, but is somewhat iffy for more numerous and variable shopper relationships.
Retailers can certainly gain some useful insights from applying RFM to their customer base, but there are several caveats to consider in formulating actions.
To begin with, RFM it is not the only relevant kind of segmentation. Separating shoppers into spending level deciles is one type of "first cut." They may also be divided by share-of-wallet, promotional responsiveness, trip missions, and their demographic similarity to other high-value shoppers.
Even if we rank shoppers by total net profit dollars, as Ryan wisely counsels, we must consider what relevant shopper and environmental traits may be masked by this metric, as well as how we can identify shoppers with the potential to be increased in value.
And finally, we must decide whether to apply these analyses globally across the entire chain, by store cluster, or on a store-by-store basis. This presents a trade-off between useful insights and operational complexity at store level.
So in short: Shopper segmentation = good. Simplistic mechanical methods = not so great. Putting insights to work in stores = pretty darned hard. Which is why ranking shoppers into sales deciles is a practical initial step on a long journey to shopper-centric retailing.