Customers on the LineQ&A with Jon Zimmerman
|
As director of Expedia's customer analytics group since its inception, Jon Zimmerman is in the unique position of having observed his company's analytic strategy in progress from its inception to the present day. We spoke with him recently about that evolution.
IE: Can you explain the genesis of your group at Expedia Inc.?
ZIMMERMAN: Sure. The group was founded in 2000 for two reasons: First, we had been collecting customer information at Expedia for quite some time but weren't doing much with it. Second, we realized there was a lot of customer information we weren't collecting at all. At that point, we determined that it would be important to start doing a good job of customer interaction management.
For example, back in 2000 as the business was starting to scale, we realized that we were getting many customers who wanted to contact us through multiple channels, including email and telephone, in addition to interacting with us through Expedia.com
But we didn't have much of a feel for what was generating support requests and whether effective steps were taken to correct perceived problems on the site. Furthermore, as Expedia continued to evolve and we moved into more product lines, our Web site grew in complexity, so it became necessary to start to understand, using real data, how people were interacting with it and where we could make improvements.
IE: How would you define "customer interaction management," as practiced at Expedia?
ZIMMERMAN: I don't know that we've ever used the term customer interaction management per se. But it's very close to describing the strategy we defined when we started the effort, which is to maximize profitability revenue and customer satisfaction by implementing processes and technologies that supported coordinated interactions.
IE: Your architecture developed in an interesting way in response to those business needs. Can you explain that process?
ZIMMERMAN: The easiest way to think about it is to look at what happened back when we started this effort. We put together a customer database, which differed from the way we traditionally stored data in a data warehouse in that we did so from a customer perspective, not a transactional one. That logically moved into an effort to do a better job marketing to those customers, at which point we realized we weren't necessarily getting all the data we needed from each customer touchpoint. So we started to think about which touchpoints were most important to get additional data from, and then once we had that data, we could determine what we needed to do at each of those touchpoints to optimize interaction with the customer.
IE: Can you give me an example of how your business has improved operationally by having that data available?
ZIMMERMAN: Sure. The main thing we wanted to do was ensure that we could support customers hopping across channels. We wanted to use information we collect at other channels to do a better job with whatever channel this customer decided to come in on.
That was a mouthful, but here's a good example: Nearly 100 percent of our customer interactions or purchases occur on the Web site, yet a very high percentage of our support occurs through another channel. So, when customers contact us via email or the telephone, it's crucial that we know how they interacted with us via the Web site. We need to make sure that call center agents have current information about customer accounts and their recent transaction history.
IE: How has your strategy evolved in the last two years?
ZIMMERMAN: The big difference is that we went from collecting the data and creating a strategy for using data to drive the business, to very carefully picking applications that lend themselves to modeling and quantitative analysis. Over time, we're ticking those applications off one by one, and along the way taking the time to stop and measure how the models are working to make incremental improvements. As we've done that, we've expanded the number of places we're using predictive modeling and other forms of quantitative analysis.
IE: Can you give me an example?
ZIMMERMAN: We're using analytics, for instance, to balance all the levers available with outbound email to drive the highest conversion-to-purchase: what offers you put in the mail, what the mail itself looks like, who the audience is, and what the timing is.
Over time, we've used analytic models to examine the data coming back and then started to refine them in terms of timing, what the particular offers are, or who the audience is.
  
|
|
|
|
|
|