Making Decisions with Bad Data

Decisions should be better with more data, but what if bad data is used to make those decisions? For Amazon, the problem was fake data. Third-party sellers on Amazon were paying people to post positive reviews even when they hadnt used a product. These fake reviews raised product ratings to lure more customers. Amazon had to spend considerable time and money correcting this problem by verifying customer reviews.⁹⁹

Outdated data is another form of bad data. United Airlines made forecasts of seating demand based on data they collected years earlier. It turned out that the data was based on assumptions about flying habits that werent true anymore. United figures it lost $1 billion in missed revenue from this mistake.

Other companies might transfer data from one part of the company to another with problems arising as a result. For example, prescriptions sent from one hospitals neo-natal intensive care unit to the hospitals pharmacy were not interpreted correctly. The pharmacys data system hadnt accounted for babies so young. This resulted in too much medicine being prescribed for them.¹⁰⁰

Data also might be collected in a companyabout customer complaints, for instance. But if that data is not shared with people who take action on those complaints, it doesnt do much good.

A recent study found that 77 percent of the 800 bankers surveyed said that their bank is not well-prepared to deal with bad data.¹⁰¹ More than half of this same surveys respondents said that their banks dont verify the data they are using to make important decisions. Results like these are sounding the alarms that data used to make decisions better be good.

Discussion Questions

1. 2-10. What should managers and companies do to minimize their use of bad data?
2. 2-11. How might intuition, the analytical decision style, and the conceptual decision style help to work against problems arising from using bad data?
3. 2-12. What does this case illustrate about big data and analytics?