Loyalty cards and the world of data mining behind them

by Guruprasad


Why do Supermarkets & retail chains offer loyalty cards to regular customers?

Should we be thankful for their generous goodwill gesture?
Why not? Afterall, we are their “privileged customer” and it is offered as a “token of appreciation” for our loyalty towards them. When the store is “generously” dishing out freebies & discounts, how can we even think of questioning them?
This is how most of us think, but there are a few who exercise logic and deduce that this is one of the ways in which the store is trying to retain (enforce?) loyalty of a customer.

But the reality is different. Loyalty cards are neither to show concerns/appreciation towards the customer nor to retain loyalty. They exist purely for reasons of analysis to help the stores improve their business model and reap more profits. One would ask how they make profit by giving away reward points because the whole thing looks so generous from the businessman’s part. That is the beauty of this model. For every rupee “generously donated” to the customer in the form of loyalty bonus or discounts, the businessman exracts 5 more rupees (approx) in the form of services “provided by the customer” and later use such data to further loosen the customer’s pockets and make more profit. Let me explain how.

In the subject of data mining, a concept called “Market Basket Analysis” deals with the analysis of shopping carts during checkout for patterns within the shopping carts and in relation with other shopping carts. One of the important analysis done in such modelling is called “Affinity analysis” in which purchases are analyzed to find the probability of items purchased together. For example, if 2 items like shaving gel and after shave lotion are found in a customer cart during billing and if such a pattern is found in several other carts as well, then the system assigns a higher weightage/level of associativity to those 2 products. In simple words, it means that when somebody buys a shaving gel, there is a very high probability of him buying a shaving lotion as well. Similarly the value of associations between bread and jam, milk & curds, coke & chips and so on. Once such associations are known, the store might place those items together or in adjacent shelves to remind the customer to buy the associated item as well in an impulse (A customer who buys coke without the intention of buying chips might end up buying chips if those 2 items are placed in proximity or a part of a promotional offer i.e Buy Coke and get Rs 5 off on chips) and hence boost sales. These are trivial examples but it has been found that based on demographics & timeline, even irrelavant items have been found to have high levels of associativity.

The studies mentioned above can be done at 2 levels:

1) At society/general level: This can be performed without loyalty cards because all they study are carts and transactions in general. Billions of transactions are analyzed using data mining software to realize a general model based on which stores can create general promotional offers. A typical result one can come up with from this data would be something like “Bread & Jam are usually bought together”.

2) At personal/individual level: This is where the loyalty cards play a significant role. Apart from the analysis of the carts, the system will be aware of the individual, his past purchases, his location and other details like frequency of purchases and its variation over time and the day of the week and so on. A specialized result one can come up with from this data would be something like “Customers who reside more than 2 kms away from the store purchase Jam in big bottles but not bread”. Maybe the reason is because such customers prefer to buy bread on a daily basis from a nearby store and visit this retail store once in a week to buy jams. When a store realizes this, it can tempt the customer into buying bread with longer shelf life by providing certain discounts only to him. (If the profit margin on such bread is Rs 6, the store might shave off Re 1 only to him if it can help a sale which in turn gives the store Rs 5 instead of Rs 0. They can realize this by mailing bread discount coupons only to those who live 2+ km away from the store).
For example, there has been a recent controversy about how a popular retail chain (Target) was able to figure out that a teenager was pregnant even before her relatives based on her purchasing patterns:

As you can notice, among the two kinds of studies, the latter which deals wth customers at an individual level is invaluable data which might be impossible to obtain even with specialized research teams as consultants. For example, a store might hire a research team and pay them Rs 10 lakh and obtain results which in turn can be used to boost profits by Rs 20 lakhs. But why spend Rs 10 lakh on research teams when your own customers are willing to give you that data for just Rs 1 lakh (assuming Rs 1 lakh is your overall cost of giving away reward points and such discounts) to boost your profits by Rs 20 lakhs with the help of loyalty cards? Moreover, the customers directly participating in these studies provide them with real data at real time which none of the research teams will be able to match inspite of their high costs and approximate predictions (Recollect that exit polls conducted by analysts are usually much farther from real results)

Now, with the above awareness, should you refuse to accept/use loyalty cards? If you are concerned about your privacy, then dont, because every purchase of yours is tracked and you might be subconsciously manipulated into purchasing goods which you might not really need (If they send you customized promotional coupons). Your records might even be sold to third parties who will use it to promote their services based on your buying habits.

However, if you are aware of these and ensure that you purchase only what you need at your freedom without being restricted to any particular store and do not fall for psychological traps, then continue using loyalty cards by all means if it can help you save money.
However, be aware of the fact that the store is not doing a favor to you by providing such services but it is the other way round. You might be doing a favor by providing services to the store at a cost much lesser than what you are getting in return.

If your brain works as a robot, with mathematical calculations & emotionless purchases, these loyalty cards will help you greatly but if your brain works as that of a typical human and is susceptible to impulse purchases, then you might have to rethink your views on these loyalty cards..

This entry has been uploaded to facebook as well followed by discussion threads with netizens:


“Why we buy” book:

Market Basket Analysis:

An interesting case study using the analysis of a fruit product:

An article on how loyalty cards and data mining are getting bigger by the day: