I started working with Joe Strickland on measuring the activity of the website created for merchants to process their transactions online. In the beginning, we had just web data and any effect on Call Center usage was hard to prove, hard to measure, and hard to improve. We realized that to effectively use the website to reduce Call Center costs, we needed a complete customer experience view. We had to integrate all information on customer touch points - web, Call Center, and satisfaction survey data. To make a very long and difficult story short, Joe identified over half a dozen systems that contained the necessary data, arranged for data feeds, and established the appropriate access. We then analyzed the data and realized that there were inherent data anomalies that required business rules in order to make data integration meaningful. For example, there was a many-to-many relationship between user ids and account numbers which had no perfect solution. Together, we came up with a set of business rules to solve these and other issues. Finally after quite a bit of "blood, sweat, and tears", we had an integrated data source that had a very complete picture of how merchants interacted with American Express.
But that was not enough. When we looked at our new integrated database, there were no clear insights of what could be done to improve merchant service and thereby merchants' satisfaction. Averages hid more than they revealed. So we started to segment merchants based on their use of both the Call Center and online services (OMS). Combining statistics and an understanding of merchants, we discovered three very actionable segments. The first segment, the segment that produced most of the Call Center calls, consisted of merchants who never used OMS. Most of these merchants only called once and called to set-up their American Express account. Nothing that Joe did to OMS would or could have any effect on these merchants without an offline strategy to get them to use OMS to set-up their account. However, it was essential that before any effort was made to get them to use the web to set-up their account, there needed to be an easy to use online account set-up process. The question was whether such a process existed. Since we had integrated satisfaction survey data with Call Center and OMS data, we were able to find the answer. The answer was no. While not the answer we wanted to hear, it at least let Joe know that there was no point in having a campaign to get new merchants to set-up their account online until the online process was improved.
The next two segments were those of merchants who used both the Call Center and OMS. Interestingly, we found that merchants who used OMS prior to calling the Call Center made significantly less calls to the Call Center. In addition, we discovered merchants who used both Call Center and OMS could be divided very nicely into two very different segments. One segment had low usage of both the Call Center and OMS and the other had heavy usage of both. Each segment required a different strategy. With closer examination of their online behavior, we found that light users were often confused when they got to the OMS homepage and didn't know where to go were often going to more pages than they really needed. To solve this problem, Joe initiated a re-design of the homepage to better direct light users to where they need to go. His re-design not only provided better routing of light users, it provided information right on the homepage so that they some merchants didn't have to go anywhere else. This new, optimized, homepage has been in full production for just over a month and has been a meaningful initial success in changing merchant behavior. Landing page optimization, while usually thought of as a web only process, had a beneficial effect on the Call Center. We are now monitoring the behavior of this group both online and offline to see how this landing page optimization works not only in the short term but also over time and when it needs to be "re-optimized".
Heavy users of the Call Center and OMS, were not affected very much by the changes to the homepage. They didn't need better routing. They were heavy users and knew where they needed to go. A different strategy was required. By examining their online and offline behaviors, we were able to determine what concerned them the most. An initiative is now underway to address these issues.
This is just first phase. In analyzing the data we've found that different industries use the Call Center and OMS in different ways and that these industries fall into a few very definite groups. We are now doing a cluster analysis to determine more precisely what parts of OMS are used by which group so that a strategy can be developed to even further improve merchant satisfaction.
Looking back over this effort I am continually reminded that in order for an effort like this to work there are two essential elements. One, the web owner must be committed to seeing the effort through - there will be many obstacles to overcome. Two, the team doing the work must not only have strong customer centric analytic capabilities, but must also have an in-depth understanding of the customers whose behavior is being analyzed. Semphonic provided the analytic capabilities and Joe Strickland and his team provided the customer understanding. Together we made it work and we've only just begun. Can Website Have an Effect on Call Centers? Yes They Can. If you do it right.