Of all data, purchase data generally represent the absolute peak of value. After all, nothing tells a marketer what a customer wants than what she has already bought. Moreover, many offers flow naturally from specific purchases, such as laptop cases for laptop buyers.
Let me introduce to you the counter-argument:
Seats 7, has 83 cupholders and can change your perspective on marketing data
Green minivans tell another story about purchase data, one that should give a smart marketer pause before blindly trusting them.
First of all, I owe a debt for this parable to a former colleague, Robert Craven. I won’t tell you how long ago he taught it to me, but the very fact that a) it was when minivans were popular and b) you could actually buy most new cars in some shade of green should tell you something.
The parable goes like this: a woman goes to a car dealer to buy a green minivan. She likes green, you see. The dealer, however, only has four white minivans and zero green ones. Now, if that dealer has any skill whatsoever, he will sell that woman a minivan--a white one.
At the micro level, this transaction results in one slightly dissatisfied customer. However, bigger things happen at the macro level. Namely, all the manufacturer knows is that white minivans are selling. The following year, they will increase the percentage of vans painted white because that’s what it looks like the consumer wants.
In effect one white minivan snowballs (pun intended) into faulty data assumptions, based on purchase behavior.
To give another example, I point to hotels. Sophisticated hotel marketers practice what I call the “I know what you did last summer” approach to offers. They analyze past stay behavior and model what location will interest guests next.
I even know that you ordered your room-service hamburger medium well, you miscreant!
Here, we have a slightly different problem than the green minivan problem. If you’ve never worked with hotel marketers, you might not know that a) most stay behavior comes from a loyalty program, which in turn means that b) most stays are for business and that c) most redemptions go to leisure properties.
Thus, let’s imagine that a business traveler has a client in Omaha. He flies out there twice a month, so he racks up a lot of stays there. Guess which destination most models would indicate as the most likely next stay? (I’ll give you a hint: Warren Buffet lives there) Correct. Now guess where the business traveler probably wants to go the least on his next vacation.
What can a marketer do to ensure that he or she uses purchase behavior data correctly? Three consideration:
- Understand what the purchase behavior really tell you. Obviously, it tells you that a specific person made a specific purchase. But consider what the information does not tell you, namely why someone made that purchase.
- Establish the why with other information. Any marketer can follow up a purchase with a survey to gather information. In the automotive category, third parties such as J.D. Power may be able to supply that information.
- Test. As always, testing hypotheses in direct communications (email, social, mobile, even direct mail) will establish which data matter and which don’t.
Any other suggestions? Please add them in the comments. And may all your minivans be green.