Data, can you dig it?

Five tips for making the most of the data you mine

by Geoff Griffin

John Wanamaker, the father of modern advertising, once said, “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.” If Wanamaker was around today, he’d have no excuse for not knowing the effectiveness of his ad dollars because of the availability of data mining. BYU professor Jim Brau points out the new standard: “As a CEO, you need to ask yourself: Do you know how effective your marketing is? Can you put an objective number on the value you are getting for what you are spending on marketing?”

Thanks to all of the information that businesses are able to collect and store in databases today, CEOs can have their customer base sliced and diced into segments that can be targeted, sales pipelines can be predicted, catchment areas for brick-and-mortar stores can be analyzed, keyword searches for online businesses can be focused and the cost of maintaining customers can be tracked down to the penny.

Data mining has become an important way for businesses to improve their efficiency. The more targeted area of Web analytics gives companies with an online presence a way to glean information about the ways in which customers use a company’s Web site.

Humans and numbers
The irony is that amid all of the computer programs and number crunching, creating a successful data-mining project involves people working together while making some very subjective decisions. John Mellor, senior vice president of business development and corporate strategy for Omniture, notes that successful data analysis requires one key component: “It takes a human.”

While data mining and Web analytics can be of great help to a business, leaders in the industry say there are certain processes to follow in order to ensure a successful project. Consider these:

1. CEO buy-in required: If the CEO or one of his or her designees is not fully invested in using data mining, the project can easily get bogged down or off track.
BYU professor Christophe Giraud-Carrier explains that employees beyond those who run the computers have to be involved. “You go to the IT department and it never goes anywhere,” he says. “Data mining is seen as a liability because it’s one more job they have to do. It’s somebody coming in and saying, ‘Here are the problems.’”
Mellor emphasizes, “The biggest thing is that there has to be a resource dedicated to it. It needs to go on somebody’s job description, otherwise it just bounces around.”


2. Consider a team approach: Hiring an outside consultant to run the numbers is just one aspect of a successful project. In-house marketing departments, advertising agencies and any other employees who can contribute should be involved. Some data miners, such as Paul Lima of Lima Consulting, even have contracts to work with different advertising agencies.

3. Decide what data you want: While information is easily obtainable in a variety of areas, the challenge comes in making the distinction between what Giraud-Carrier refers to as “rubbish data vs. quality data.”
Lima recommends a team approach in “understanding the problem you’re trying to solve. Don’t overestimate the simplicity of the problem. You may need to meet for hours. You need to come up with a one-sentence objective that states the problem, not the solution.” Only then are you ready to start a data mining project.

4. Be ready to act on what you find: “It’s very easy to collect information,” notes Mellor. “The value comes when you use it to measure and test.” Numbers are of little value to a company that is not prepared to learn from them.
“The key to success is if people are prepared to learn, it’s always successful, even if you don’t end up changing your business model,” says Giraud-Carrier.
Lima warns that companies also need to beware of the belief that data mining will provide some sort of silver bullet. He instead predicts “marginal improvement across a wide variety of areas” will add up to a better business model.

5. Keep a positive attitude: When sorting through that daunting pile of statistics, Mellor advises to keep in mind, “There’s a pony in there somewhere.”
In the end, the belief that data mining can help your company can be just as important as the hard numbers you find.

Number-crunching Cougs

BYU's data mining lab allows companies to dip into the data world without a big committment

Data mining has become such an important part of business that Brigham Young University has set up a data mining lab, where students can get experience crunching numbers and applying them to real-world situations on behalf of companies that use the lab’s services. The lab is part of BYU’s computer science department and typically has six to eight undergraduate and graduate students working under the direction of Professor Christophe Giraud-Carrier, who opened the lab in 2004 and who also teaches courses in data mining.

“We work with industry partners to help them find things they did not know about,” explains Ph.D. candidate Matt Smith. “It gives us an opportunity to work with real data.” Many students who have worked in the lab have gone on to data mining positions with companies such as Google and Amazon.

Giraud-Carrier says, “We’re not a commercial interest,” so companies will generally reach a “gentleman’s agreement” with the lab wherein a donation is made to the school to help fund the students who work on the projects.

The lab also offers the chance for companies to dip their toes in the data mining waters to see what it can do for them without having to commit major resources right off the bat. Giraud-Carrier advises, “Start small and build from there. Brainstorm about a problem and use a consultant on a two- to three-week project in a restricted area. You can build from there to success.”
– GG

Defining data mining

Just how exactly does one define data mining? Is it the same as Web analytics? Some say data mining occurs any time you take any amount of data and use it to determine how to better your business. Others suggest it only deserves that title if higher-order math, such as regression analysis, is involved. Is data mining a tool or a process? Should it be conducted by an outside consultant or managed as part of a multi-disciplinary team approach?

The whole area is still cutting-edge enough that the exact definitions haven't yet been worked out. For purposes of the information on this page, let’s just call it all of the above, with the caveat that Web analytics specifically applies to data derived from business Internet sites.
– GG