Answering the real questions in data analysis

A guy walks into your cube and asks you to whip up an econometric model. You’re a statistician, after all, and you’ve got a Ph.D. in something or other. You do this for lunch, he figures.

He “over-thought,” says the one whose cube such a guy walked into. Theresa Doyon has been routinely navigating datasets in the 50 to 300 million-unit range for 10 years. She’s good at all-terrain tools like SAS and KXEN. She could have produced the report he wanted. Instead, she asked him, “Why?”

They talked for an hour. What he actually needed, she discovered, was more of a descriptive report — something that gives a picture of a current situation. All he wanted to do was figure out how to allocate resources.

This is where things usually go wrong. “We’ve got all these great tools,” she says, “but I don’t see us using them as well as we could.” We could write this off as a communication problem, but she believes it’s much more than that.

It’s something like a man who walks into a kitchen with a carton of eggs and tells the cook to fry them all — when actually he’s just hungry and happened to have found eggs. Something else might suit him better. Bacon?

Business people complain that they get reports, not insights. But they’re not sure how to ask for the data. Meanwhile, analysts complain that they’re asked for a fire hose of analysis and that it’s always due yesterday. But they deliver. Asked for a report, they produce a report. Asked for a model, they produce a model.

“But that’s not what the business people really want,” says Theresa. ‘What they really want is to answer some sort of business question, like how’s my marketing doing? What can I do differently?”

“If you had a strategy and you did bite-size tests and learned as you go,” she says, “you could start to use the analytics that would really drive the insight.” We can really improve the way things are done. That’s the missing link.

At least some organizations have found that link. A home furnishings retailer she worked with recently had her work closely with marketing people as their questions and her analysis evolved.

The retailer had been suffering as ever more of its customers feared they’d soon lose their homes. The living room sets that looked so good just months before seemed to lose their appeal. Theresa’s assignment was to, in effect, come up with a silver lining.

“It was a very big and ambiguous question,” she recalls. She worked with the client on a series of projects over nearly two years as the sour economy evolved. She estimated opportunities, customer targets, and gave the marketing people she worked with critical guidance on the launch of new programs. She recalls, “It came out quite well.”

See her LinkedIn page here.

One Response to Answering the real questions in data analysis

The data industry thrives on conversation. Please submit a comment.

Other recent posts

Bohemian Grove a la BI

The Bohemian Grove of the BI industry convenes for the fifteenth time in just three weeks. Naturally, you ask the obvious question: Are you serious? The Grove? A summit? The answer begins with a fond recollection of the Grove. If you’ve never attended the Bohemian Grove yourself — I haven’t, though I live in the… Continue Reading

Favorite Star Trek, a data story

This story shows how elemental data stories really are. Humans come ready to tell and hear them, requiring no plug-ins at all. This young person can do a good job of it. There was a question, followed by data, then questions and answers, and and finally a conclusion. It’s all there. It’s elementary. Sure, this… Continue Reading

Bad stories stop good data at the water cooler

We agree by now that data’s a good compass. One neglected question is tougher: Which map? Everyone’s known the kind of “grouchy guy” TDWI instructor Kellee M. Franklin, Ph.D tells about. This guy knew better than most of his co-workers about how their Washington, D.C. defense agency worked. And he was frustrated. Over the years,… Continue Reading