News Stay informed about the latest enterprise technology news and product updates.

Five tips for firing up a BI analytics practice, and some reality checks

Here are five tips for getting a predictive BI analytics practice off the ground.

KISSIMMEE, Fla. -- The BI enthusiasts at this week's Information Builders Summit made predictive BI analytics seem like the next best thing to a time machine or crystal ball. A data-driven crystal ball, that is.

Predictive BI analytics, like many an IT term, refers to an amorphous discipline that involves using statistical and other complex mathematical analyses of data to predict the future. It is not exactly data mining, the industry experts like to point out, and it's certainly not data mining for data mining's sake. Predictive BI analytics identifies patterns in the volumes of data generated by a business and by relevant external sources. Those patterns are used to predict the future so the business can adjust its strategies accordingly.

Using analytics to shape business strategy is not easy to do, however, and not just because it can be expensive or because it's a relatively new undertaking for many but the biggest corporations. Putting predictive analytics to work in the corporation requires change management skills, finding the right people and the right mind-set.

Here is a sampling of tips from Wayne Eckerson, director of research at TDWI Research in Hingham, Mass., whose session at the conference dealt with what CIOs should keep in mind as they get started on a BI analytics program -- and some reality checks from CIO attendees.

  1. Find great analysts. People good at this flavor of predictive analytics not only are inquisitive and critical thinkers but also like to experiment and are "doggedly persistent." They have to know both the data and the business processes that produce the data inside out; they also need to know how to use the tools -- a rare combination. The best analysts are not philosophers living in ivory towers; they know which business questions to ask for actionable results -- increased profits, for example, Eckerson said.

    Candidates nowadays can come from MBA programs, Eckerson said. Social scientists, statisticians and people trained in Six Sigma have an affinity for this work, as do ambitious data analysts ready to take their career to the next step. The best people are pricey. Competitors will try to woo them away.

  2. Put analysts into a centralized group. These analysts will work better and more creatively if they are together, Eckerson believes. So, rather than "bury" them in business departments, he suggests that companies centralize business analysts and locate them near the data warehousing team, the people with whom they will be working closely.

    In answer to a question from a conference attendee about to whom analytics people should report, Eckerson said he has seen analysts reporting to someone outside the IT department: the chief operating officer, a financial executive or even the CEO.

  3. A reality check for smaller companies: Conference attendee CIO Chris Brady of Carmel, Ind.-based Dealer Services Corp., an $150 million inventory lender for car dealers formed in 2005, chuckled when she was told about these hothouse predictive analysts. Her company successfully used predictive analytics to reduce risk and help its car dealers better manage their inventory during the recession. But she doesn't have anybody with the title "business analyst." "We did this in addition to our usual roles," she said. "But I agree it takes people with a thirst for knowledge, who really want to look at this stuff."

    Never come off smarter than the executive you're supporting, or suggest the data model is smarter than the executive.

    Wayne Eckerson, director of research, TDWI Research

    Conference attendee J. Ed Smith, CIO of Hanover, Penn.-based snack food maker Utz Quality Foods Inc., can attest to the importance of identifying and nurturing analytics people. Utz has been using Information Builders' WebFocus BI platform since 1998, and is looking for ways to deploy it across the company. Smith brought a help desk technician to the conference with him to learn more about BI. The technician, who has shown promise in analytics, is cutting his teeth on a project that analyzes Utz product profit margins.

  4. Cultivate a fact-based culture. A fact-based decision-making culture is willing to test assumptions, embraces transparency and often uses dashboards up and down the organization. Such organizations also recruit other analytical leaders. The very top leadership helps decide which analytics projects to take on and when, and funds them. Executive support is critical because the best analytics projects cut across departments, Eckerson said.

  5. Reality check: If the leaders of your business are not "show me the data" types, with an appreciation of "hands-on" experimentation, "you may be out of luck," Eckerson told the audience. He suggested going elsewhere, or waiting. "If your company is not analytically inclined, it may start to have problems, and new leadership will be brought in."

  6. Start by testing one assumption to gain support: Predictive BI analytics is about testing assumptions, which by definition are hard to dislodge. By starting small, with a pilot project, it is sometimes possible to effect a culture shift.

    An example: To gain support for analytics, a BI professional at a major retailer of online office supplies asked to test a long-held assumption of the business: the belief that online customers stopped ordering or decreased their orders if a big office supply store, like a Staples, was located within a certain distance of them. In fact, the data showed that the critical factor was not geography, but frequency of purchases. Online customers who placed four orders every 60 business days showed a retention rate of 95%, regardless of the number of stores on the ground. Frequency of purchases is a key metric for measuring customer retention, and the sales exec immediately "got it," Eckerson said.

    That success story allowed the BI professional to take on a bigger data modeling project, and it brings us to Tip 5.

  7. Don't get uppity. Never come off smarter than the executive you're supporting, or suggest the data model is smarter than the executive. "It is the kiss of death," Eckerson said, echoing a warning voiced in many sessions at the show. If the results from the data modeling and statistical analysis make "intuitive sense" to the business exec, as in the example above, so much the better.

Dealer Services' Chris Brady said that not all predictive BI analytics projects have to be massive. Her company uses WebFocus RStat for quick correlations that can be equally effective in changing minds.

Brady offered one hypothetical that a quick correlation could easily prove wrong -- and change minds: Forty percent of automobile dealers have blond hair, and all blond-haired dealers default. "I have heard statements like that pronounced as fact," she said. The last five defaulters with whom that business person dealt indeed may have been blondes, but it doesn't make the assumption correct, she said. "It is not hard to take that data and do a correlation graph showing that blond hair does not correlate to default."

Let us know what you think about the story; email Linda Tucci, Senior News Writer.

Dig Deeper on Enterprise business intelligence software and big data

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.