For online processing sites, the integration of Call Center and web usage data is a powerful tool to improve customer satisfaction and reduce costs. Call Center support is much more expensive than web support, but there are some things that need the human touch of a live person. The goal is to maximize the use of the web for customer support without harming customer satisfaction or even worse losing a valued customer. By understanding how customers use both the Call Center and the web and the interaction between these two support channels customer support can be optimized - high satisfaction with low costs. Call Center data is heavily reliant on human judgment. Web usage data is reliant on effective implementation of technology. While there are technology issues in integrating the two sources, the issue of integrating human judgment with technology is even more difficult. The ability to use technology to improve human judgment is critical.
There are several challenges in integrating data from these two very different sources. Assuming that data is being collected for both in an electronic format that is able to identify the customer, there has to be a common key that connects a customer call with customer online usage. This can be an account number or a user id. In many companies, Call Centers and web sites use a different identifier. Call Centers may use an account number and the web uses a user id. If there is no table that ties account number to user id, then there is no way to integrate Call and web data. If there is a one-to-one correspondence between account number and user id or if both systems use the same identifier then there is no problem. Often reality is somewhere in between. For example, sometimes there is only a partial matching between Call Center and web identifiers. In other cases, there is a many-to-many relationship between the two identifiers. An account number is associated with multiple user ids, and a user id is associated with multiple account numbers. To solve these in between situations, companies must develop business rules that can be used handle ambiguous situations. These can range from ignoring ambiguous data and keeping track of how much data is ignored to grouping data together to eliminate the ambiguity. The grouping of data often happens in the financial industry where family members have multiple accounts and multiple user ids which can access these accounts. To solve this problem, accounts and user ids together that belong to single family are grouped. This technique is called "Householding", but it is not perfect. Sometimes it is difficult to determine who belongs to a single "household". Sometimes the difference between family member behaviors is significant and Householding hides important insights. Whatever the solution, the business rules must be well articulated and the consequences of their usage understood and accepted. As with all data integration projects compromises are essential. Don't let the perfect get in the way of the good.
Once a common key is established, the challenge of data compatibility must be addressed. When a call is made to the Call Center, the operator can record a wide variety of information ranging from call type to call resolution. Similarly on the web, a customer can view different content and perform different functions. For the integration of Call Center and web data to be effective, the information recorded by the Call Center must relate to the information collected on the web. For example, if a customer can change account information either through the Call Center or through the web, both systems must classify that action similarly. The closer the classification, the better.
Having a good (though not perfect common identifier) and matching classifications are the essential foundations of a good integration of Call and web data. It is not enough. Good Call Center data is dependent on the input of the operators. If the operators are not properly trained, the data they input is bad. Call Center training, turnover, and skill level are a never ending challenge. It turns out that understanding how Call Center reps respond to customers can be critical.
This leads me to the meaning of "Other" and why it is so important. At first, we ignored calls classified as other. Then one day instead of looking at "Other" in total, we looked at "Other" calls on an individual customer basis and found a very interesting pattern. Most customers who called had a small percentage of "Other" classified calls. However, callers who had a sudden increase in the number of calls had most of their calls classified as "Other". Then when the number of calls dropped down to a normal level, the percentage of "Other" calls returned its lower level. We then combined satisfaction survey data to the Call Center data, we found that the customers who had a sudden increase in calls most of which were classified as "Other" were also very dissatisfied. Angry, dissatisfied, callers consistently had their call type classified as "Other". This was very useful information. It was as if we had a new call type: "Angry". We could track the number of customers who had a high number of "Other" calls and know how many angry customers there were. The only thing we didn't know was why they were angry. The integration with web data provided the answer. We looked at web usage for the customers during the period when they had a high number of "Other" classified calls. In almost every case, for each angry customer there was a corresponding increase in site usage for a particular area of the site. Keeping track of what site areas makes customers angry, is a very good way to prioritize web improvements. It also helps create a strategy for how to improve Call Center training.