Site icon IT World Canada

7 ways to avoid the traps of big data

There is this often repeated story of how the big box chain Target Corp. was able to determine that a teen aged girl in the United States was pregnant even before her father knew about it.

Target caught the parent’s ire when the store started sending her coupons and ads for baby products. The manager rightly apologized, but days later, the story goes, he receives a call from a now more apologetic father informing him that the girl was actually expecting.

Target uses a pregnancy prediction scoring system which infers from past purchases of a female customer what might be the best time to send this person targeted baby product ads and coupons.

To some retail businesses the story may illustrate how today’s data driven analytics tools enable organization to use seemingly insignificant information to carryout profit boosting campaigns.

However, Ikhlaq Sidhu, chief scientist of the Coleman Fung Institute for Engineering Leadership, cautions that it doesn’t matter what big data analytics tools a business may have, it’s bound to get in trouble if it fails to back it up with sound judgment.


“Without a skilled pilot, the new technology is more dangerous that it is helpful,” he wrote in a recent blog for the University of California Berkeley online publication, The Berkeley Blog.

Sidhu outlined seven key lessons that businesses can learn from the Target experience.

First of all, he said, businesses need to have a “revenue-driven” or “risk management-driven” business case for using big data.

He said Target was acting on the business thesis based on studies from the 1980s that people who “mindlessly” buy everyday items by habit are immune to advertising or coupons designed to get them to switch brands.

There is an exception to this rule, however. When life changing events such as getting married or having a child occurs, purchasing habits are likely to change.

Sidhu said without this business insight, Target’s pregnancy prediction score and ad targeting system would have been useless.

Sidhu further discusses other factors such as the need for modeling and simulation, a continuous process and most importantly, “judgment for ethical considerations and privacy.”

Read the whole story here

 

 

 

Exit mobile version