CommercialPOV

Charles Chung, Vice President of Information Intelligence, Experian

What type of advice might you have for companies implementing data mining systems?

Certainly, they need to have their eye towards the customer in collecting data, analyzing and using it. It really needs to be utilized to benefit the customer. In my area, we’re really more focused on the analysis of the data, and helping our clients mine it for making decisions. In that process, what’s really important is the quality of the data. What we preach is data advocacy within the organization. If they collect data—and they’re not data advocates—then they’re not going to be the kind of attention given to collecting the data, analyzing and utilizing it properly. We want to increase the awareness and the advocacy of the data as an important corporate asset.

So what happens when you don’t take those precautions and you plug dirty data into these systems.

Well, if the data’s dirty, then you’re going to make bad decisions. The bottom line is that you’re collecting the data so that you can make decisions using that data. If you have bad data—or it’s not accurate—basically, you’re going to make bad decisions. The idea is that if you increase the data advocacy within the organization, then you’re going to give the proper attention, the proper investment and proper resources to make sure that you collect it, analyze it and utilize accurately and relevantly. That’s certainly our focus.

Are you building metadata about large groups of customers?

We do that as well. We will develop metadata—being information about the data. We provide decision solution tools. So things like models or knowledge that you can drive from analyzing the data. Those are the results of having good, quality data. You’re then able to drive better knowledge and better decision solution tools. The metadata is a very important first step. Make sure you assess the data you have, and understand it, so you know if it’s of poor quality…you know where the holes are, and you can improve it.

What would you say to a customer who might be concerned about how a company is handling their data? Is it possible for a consumer to be lumped into a category that they don’t deserve to be in?

CC: It really gets back to the data quality issue, and I basically see five major obstacles to having quality data. One is that the data is not well documented, so people aren’t real clear on what’s all there and how to use it. Secondly, organizations tend to poorly manage data. How it’s collected may not be reviewed once it’s set up, and yet there could be something happening in the collection process where data’s not being collected properly—or once it’s collected, it may not be being loaded correctly or transformed accurately. A lot of organizations are very poor at managing the data, and some are just letting it stockpile without cleaning it up and getting relevant information. Thirdly, data is not integrated, and that’s where some of the mismatching happens—where a client who should be within a certain segment may not be, because you haven’t been able to integrate it across the different data silos. The fourth obstacle is when data is not valued. Because the data is not valued, you’re not getting the proper investment or resources to be able to clean it up, to be able to maintain it, to be able to check it. Lastly, sometimes organizations have very relevant data that would be very useful in servicing their customers better, yet not every organization within the company has access to it. Certainly, there’s data to which you want to limit access, but for the most part, client data should be available and accessible to everyone in the organization. A lot of organizations just don’t have a very efficient way of getting to data.

What about other business conditions—such as Mergers and Acquisitions?

When you merge two organizations, you bring multiple databases together. How you integrate those databases is hugely important. If you don’t do that well, you’re going to have inaccurate information within your consolidated database. To me, that spells trouble. You could have had a very loyal, very valuable client from one organization, and when they get acquired—and that individual’s information is integrated into the new database—that valuable client’s information is either lost or mis-transformed or loaded in a bad fashion. Well, now you’ve got a valuable customer who might not be given the same kind of treatment or service once the acquisition has happened. So how the data gets integrated is hugely important. Also just understanding the data. For example—going back to the documentation—when you have mergers and acquisitions, if the data isn’t well-documented, the acquiring firm is going to have a hard time integrating and utilizing the data from the company that they’re acquiring. As companies go into M&A opportunities, they have to look at the database of the organizations they’re acquiring and look at how well it’s documented, and look at the practices of managing that data. Have they applied best practices? All of that—to me—should really work into the valuation of the company they’re about to acquire.

How big is the element of human error in this? When you’re trying to draw correlation between data and tendencies…

In my mind, if you have the data, you’ve got to allow the data to speak for itself. When you substitute human decisions with a more data-driven decision, you’re going to actually make better decisions—if the dats is accurate. In fact, the more human intervention you introduce to customer relationship—I think you’re opening it up for more judgmental and irrelevant decisions to be made. As a human, you have certain individual opinions and perceptions. The data might not reflect that, but it’s going to be ingrained in your mind. I’ll give you a specific example:

My wife was very set on getting a BMW, and we had done all the research and we were pretty much at the point of going in and getting the car. When we got there, we were in our shorts and sandals, and we got absolutely no help for about 45 minutes. And when we finally did get help, it wasn’t very friendly. Right there, you’re allowing the human to make the decision around how he or she should treat the customer. But if they had data on us, they’d find that we were very credit worthy…and we were actually the perfect customer. The result of that was that my wife ended up buying an Acura. Sometimes in relying on data—as long as it’s managed well—less human error is introduced and more relevant decisions are made.