Site icon IT World Canada

Avoid these big mistakes with big data

Data Analytics Charts

Image from Shutterstock.com

All too often some companies make big mistakes with their big data implementation. When this happens, clumsy analytics, bad data and company politics are the usual suspects.

The “dirty secret” of the analytics world is that it is riddled with errors, according to the founder of what is considered to be India’s second largest analytics firm.

When businesses don’t to analytics right, they run the danger of coming up with “very wrong conclusions,” according to Srikanth Velamakanni, CEO of Fractal Analytics.

For instance, Velamakanni said, his company was building a predictive churn model for a telecom operator when they found that that some of the models they come up with appeared to be “so predictive that it was suspicious.”

Velamakanni’s team was trying to predict who among the telecom’s subscriber where likely to cancel subscription.

At that time telecom operators charged a small deposit for handsets, and he said one of the variables that were highly predictive was that customers who did not have a deposit were likely to leave the company.

It seemed simple enough, if the customer did not put down a deposit, that customer was not likely to stick around. But Velamakanni thought this was “too good to be true,” so he and his team conducted a deeper investigation.

They found out that the system used by the telecom was programmed in such a way that if a customer canceled their subscription, their deposit would “go out of view.”

Velamakanni called this an “after-the-fact variable.”

It was actually a case of cause and effect.

Looking at the data, analysts that if customers had no deposit, they were likely to leave, when in reality the deposit could not be seen in the system because it has already deleted when the customer left.

The mistake was that analysts were looking at inadequate data.

Very often data, itself could be the problem

There are many examples of companies that have been burned by bad data.

Velamakanni warned that handling huge amounts of data comes with a lot of “messiness in it.” He said very often, there can be a lot of missing value in the data collected as well as different kinds of issues that may arise when companies gather what appears to be conflicting data.

Adding company politics into the mix can also be very damaging he said.

Analytics can be used to prove a certain point but this could be a “very faulty way of coming up with analysis,” he said.

For instance, Velamakanni said, his company has had to deal with some clients to don’t was to democratize analytics inside their business because they fear it could be used by staff to justify conclusion that they deem as right.

Read the whole story here

 

 

 

Exit mobile version