Big Data Fails: How to Avoid Them

Big Data: 6 Real-Life Business Cases

Here's how the implementation of big data initiatives typically goes at big companies these days: An organization deploys Hadoop, hires some data scientists, and waits for magic to happen. But the magic doesn't happen.

That's the scenario described by Bill Schmarzo, CTO of EMC's Global Services big data practice, and author of the book Big Data -- Understanding How Data Powers Big Business. Schmarzo also teaches the course Big Data MBA at the University of San Francisco.

Schmarzo's group at EMC doesn't sell products. Rather, it helps large corporate customers figure out what to do with their big data investments. He's seen many stalled big data initiatives. His group comes in when the stall happens. What's causing all these stalls?

"Organizations don't need a big data strategy as much as they need a business strategy to do with big data," Schmarzo told InformationWeek during an interview at Strata + Hadoop World in New York City this week. 

[Looking for more about what happened at Strata + Hadoop? Read Cloudera Boosts Hadoop Portfolio With Security Data Update, Offerings.]

When Schmarzo's group starts working with a customer organization, the first step is to figure out what the business is trying to accomplish in the next six to nine months.

"We have users that have these hunches about how things work because they've been living in a business environment for so long," he said. "But what are the business decisions they are trying to drive off of that?"

For example, Schmarzo tells the story of a hospital in a major city that saw an increase in injuries after a local team's sporting events. The people who worked there recognized the correlation. But how could they measure it and use the information to make business decisions? Schmarzo said that an investigation of the data showed that injuries went up by 27% on game day, and that quantification of the injuries was something the hospital could use to figure out how much it needed to augment its ER nursing staff for game days.

In another example, Schmarzo's team helped a casino that was trying to understand how to get more lifetime value from casino visitors. The company typically offered high-roller rooms to visitors who gambled a lot in one day. A big data analysis examined what would happen if the casino offered those high-roller rooms to other visitors. The result was that other visitors who were offered high-roller rooms ended up staying even longer and spending more than the high rollers did. They were potentially worth more to the casino in the long run.

These are a few of the success stories Schmarzo has seen in the EMC practice. But there are still plenty of challenges for organizations implementing big data initiatives. And the biggest one is cultural, not technological.

Schmarzo said his group sees the most success with midmarket companies. And there's another factor that seems to drive the success of these projects.

The companies that run into the most trouble are those in which data is in silos, and the thinking about that data is also in silos. For instance, in a banking company there may be a checking account silo and a mortgage silo, and the owners of each group aren't accustomed to thinking about the whole customer who consumes both services.

(Image: NorthernStock/iStockphoto)

Companies that can get past that limitation in their thinking are more likely to be successful with their big data initiatives.

And that example also shows an important factor in successful big data initiatives – collaboration among groups who may not normally collaborate with each other. It relies on team members with different areas of expertise working well together.

"The places where we are seeing success is where the business people and the IT people like each other," Schmarzo said.