By using website you agree to our use of cookies as described in our cookie policy. Learn More


TDWI Blog: Data 360

Blog archive

Art, Science, and Analytics

It’s easy to get mesmerized by analytics. The science behind it can be intimidating, causing some people to abandon common sense when making decisions. Just ask financial executives of major investment houses. Blinded by complex risk models, many took on too much debt and then faltered as the economy tightened in 2008.

Relying too much on analytics is just as disastrous as ignoring it and running the business on gut instinct alone. The key is to blend analytics and instinct—or art and science, if you will—to optimize corporate decision making. Interestingly, several business intelligence (BI) leaders use the phrase “art and science” when discussing best practices for implementing analytics.

Stocking Auto Parts

For example, Advance Auto Parts, a $5 billion retailer of auto parts, recently began using analytical models to help it move from a “one size fits all” strategy for stocking inventory at its 3,500 stores to customized inventory for each store based on the characteristics of its local market. This store-specific assortment strategy has reduced non-working inventory from 20% to 4%, generating millions in cost savings.

“By blending art and science, we gain the best of each and minimize the downsides,” said Bill Robinette, director of business intelligence at Advance Auto Parts in a presentation he delivered this week at TDWI’s “Deep Analytics for Big Data” Summit in San Diego.

Advance Auto Parts sells about 600,000 unique items—everything from windshield wipers and car wax to transmissions and engines—but each store can only stock about 18,000 parts. To determine the best items that each store should carry, the company combines business rules based on the accumulated experience of store managers and executives with analytical rules derived from logistic regressions, decision trees, and neural network algorithms.

This blend of business and analytical rules has proven more accurate than using either type alone, says Robinette. Plus, it’s easier to sell analytics to long-time managers and executives if they know the models are based on commonsense rules that they created. And of course, the results speak for themselves and managers now wholeheartedly back the system.

HIPPOs and Groundhogs

Echoing the theme of blending art and science, Ken Rudin, general manager of analytics and social networking, at Zynga, an online gaming company, recently discussed the dangers of making decisions solely using intuition or analytics at TDWI’s BI Executive Summit in August.

Rudin uses the term HiPPO to explain the dangers of making decisions without facts. A HiPPO is the highest paid person’s opinion in the room. “In the absence of data, HiPPOs always win the discussion,” says Rudin. Zynga now examine every idea proposed by game designers using A/B testing on its Web site to assess whether the idea will help the company increase player retention, which is a key corporate objectives, among other things.

Conversely, Rudin says companies should avoid the “Groundhog” effect. This is when people focus too much on scientific analysis and data when making decisions. Groundhogs get caught up in the details and fail to see the big picture. As a result, they make suboptimal decisions.

For example, A/B testing revealed that players of Zynga’s Treasure Island game preferred smaller islands to minimize the amount of time they had to dig for buried treasure. Consequently, Zynga’s game designers made islands smaller to reduce churn, but each time they did, player behavior didn’t change much. What the testing missed, says Rudin, was that the players didn’t mind digging as long as they were entertained along the way, with clues for solving puzzles or notes left by friends, for example.

Happiness and fulfillment in life comes from achieving balance often by blending opposites into a unified whole. The same is true in BI. Successful BI managers combine art and science—or intuition and analytics—to achieve optimal value from their analytical investments.

Posted on October 7, 2010


Average Rating

Add your Comment

Your Name:(optional)
Your Email:(optional)
Your Location:(optional)
Please type the letters/numbers you see above.