Around 80% of data held within an organization is in the form of text documents—for example, reports, web pages, e-mails, and call center notes. Text is a key factor in enabling organizations to gain a better understanding of their customers’ behavior. No wonder text analytics are gaining recognition, especially in the call center industry.
It might be possible to use the insights derived from analyzing call center traffic to improve customer satisfaction, reduce wait times at call centers (improve throughput), and reduce call volumes by analyzing the causes of customer interactions. However, there are problems for which an analysis of structured data might not suffice. One such problem is identifying the main causes that resulted in a call to a call center. The reason for this is that the cause of a call is not a factor that can be easily measured. It might, however, be possible to discover the causes of a customer call from the unstructured data that is available in the customer relationship management (CRM) systems used in call centers.
This blog tries to explain the stages involved in a text analytics program. There are overall four such stages.
Identifying the issues that needs to be answered
Some of the general problems in call centers are: Why are the customers calling? Who are high-profile customers? How can cross-selling take place? Do note that these are general business statements and may not help in pinpointing the exact pain points. Such problems need to be broken down into subcategories, and then variables responsible for each of them need to be identified. It is very much necessary to understand the underlying needs and the right context. It is critical to frame problems accurately and realistically.
Identifying the best possible solution
In analytics, identifying the best solution is not the best method of approach. Each problem can be solved in multiple ways, and not all of them need to be data driven. This is because what may seem to be the best in theory may not be implementable in real-life scenarios. This could be because of the lack of good data or because of departmental silos. Also, implementing ideas at an organizational level requires the approval of all interested stakeholders. So the best solution is always the one that can be implemented!
Implementing the “Best” Solution
This is the stage that takes the most amount of time in any text-analytics program. This phase typically takes more than 60% of the overall time and effort. It begins with collecting data from all sources available. Data mining, data scrubbing, and data exploration are the next steps, which consume most of the time. Sentiment analysis and predictive modeling are what follow after that. Predictive analytics may not be needed for all the cases. Once calibrated properly, models are finalized. And lastly, the results are then interpreted and presented in the business context.
This is a step that is sometimes ignored. For any text-analytics project to be successful, implementation of the suggested solution should be successful. It is absolutely necessary to observe whether all the assumptions and constraints hold true in real-life scenarios. Any deviation from the obvious must be immediately captured and assessed.