Follow Datanami:
July 13, 2016

Secrets of the Data Whisperer


As big data technology evolves, so does the role humans play to operationalize data science. Well-defined job descriptions exist for folks like data scientists, data analysts, and data engineers, but there’s a certain class of data problem that aren’t always good match for their abilities. To fill in the gaps, some big data practitioners are turning to an emerging role dubbed the data whisperer.

What is a data whisperer? Just as the Dog Whisperer has a deep understanding of our canine friends, the data whisperer displays an innate ability to comprehend data’s role in a business. Whereas data scientists are typically immersed in mathematical models, data whisperers are experts at connecting those models to the actual business problem at hand.

Data whisperers are also highly tuned to human behavior, which doesn’t always reflect accurately in the data, or fit neatly into data scientists’ models. Whereas the scientist will scratch his head and wonder why a highly tuned algorithm doesn’t work so well in the real world, the data whisperer nods his head and understands that humans are complex and don’t always behave in rational ways.

As an executive with a company that makes enterprise contact center software, Matt Matsui understands this better than most. “Predictive analytics is not an exact science,” says Matsui, who’s the SVP of products, markets, and organizational strategy at Calabrio.

“People get hung up on math, algorithms, models, numbers and data, and the truth is all of those are just proxies for trying to predict and understand human behavior,” he continues. “That’s the part that gets lost in this so often–that all those numbers are really in service to try to predict something that’s predictably unpredictable.”

Whispering Sweet Subtleties

data outlier_1

Data whisperers help to close the gap between data models and reality

That’s not to say that Calabrio isn’t quantitative in its approach. In fact, the Minneapolis company has embraced data science and data analytics and uses the latest technologies and techniques to help its customers glean insights from data generated from its contact center software.

For example, the company predicts net promoter scores and customer satisfaction scores for its customers, and forecasts when customer service representatives (CSRs) or customers are getting ready to churn.

But even these fairly widely adopted data analytics use cases are fallible to misinterpretation.

“That idea in aggregate is fairly clean. I can give you an index score that will tell you how likely you are to leave me,” Matsui says. “But when you try to apply it to an individual customer who’s calling in, it really requires the human mind, actually, to be able to interpret those subtle signals that are spoken and not spoken.”

There are certain things people should be aware of when trying to apply models created from big aggregated data to specific people, Matsui says. “When you go to apply an insight, you don’t know what piece of data you have and what individual data point you’re dealing with,” he says. “Is that one of the medium points? Is that an end point? Is that something that falls outside of the norm?

“When you’re dealing with a specific person and you’re trying to apply a statistical average to an individual data point, that’s where the human side of it comes in.”

Enter the Whisperer



The term “data whisperer” isn’t new. In fact, it’s been used for at least the last three years –Tim Negris had a good explanation of data whisperers in this February 2013 Data Science Central article. But Matsui and others are beginning to refresh the term and bring it back into vogue.

In Calabrio’s case, data whisperers are business-oriented folks who work with data scientists to ensure that the right data and the right models are being used to solve specific business problems.

In the context of a call center with tens of thousands CSRs dealing with millions of telephone calls, that means automating the handling of every call that can be effectively automated, while ensuring that human CSRs are available to deal with the exceptions.

“If you have an agent and a set of at-risk customers, I may run the most at-risk customers a program that does an automated communication without a data center agent at all,” Matsui says. “But there’s always going to be exceptions. There will always be high dollar customers that you want to handle individually. You may want to have a data whisperer fill in the blanks and look at the overall model and make sure you’re not missing data.

“That’s the Achilles Heel of big data: you never know if you have the right data,” he continues. “As you know, it’s not just more is better. It’s also got to have a certain mix of data. It can’t all be the same kind of information. The more varied information you have, the less risk is incorporated into your model.”

Data whisperers are more human-oriented data professionals who understand the business. It could be an experienced CSR or a business analyst, but it’s probably not going to be the data scientist, who is focused mostly on math and statistical models.

The data whisperer doesn’t get flummoxed when the models don’t turn out right in the real world. Instead, he realizes that the models provide useful information, but only to a point. To really move the needle when dealing with humans, the data whisperer brings a human touch. This is particularly true when dealing with advertising, marketing, and customer service categories. (It’s obviously not true when dealing with temperature data streaming off an airplane.)

According to Matsui, a data whisperer helps to fill in the gaps as it pertains to information management. “Typically we see data whisperers in practice, they’re from the line of business,” he says. “They’re not from the contact center or IT. They’re not data savvy necessarily, but they do understand the market and what drives commerce.”

As organizations seek to leverage their big data for competitive advantage, the data whisperer’s job description is likely to change. We’re currently in the midst of a period of rapid technological and business change, where data whisperers can really help to fill in the gaps. As big data analytics becomes more pervasive, it’s possible that data whisperers will become less needed.

Related Items:

8 Tips for Achieving ROI with Your Data Lake

What’s Hot This Summer: Data Science Bootcamps

Avoid These Five Big Data Governance Mistakes