About Cisco CISCO BroadSoft is now part of Cisco Learn More


More and more organizations today are moving to unified communications (UC) platforms for better communications within their organization—with their customers and with their partners. These platforms combine voice, email, chat, and web into a seamless omni-channel experience for its users. They boast of a number of features, but most of them provide either static or rule-based experiences. Given that these platforms generate tons of data, can this data be used to improve user experiences in an intelligent manner?

Predictive Analytics is starting to play an increasingly vital role in predicting user intentions. Whether its search engines, online shopping, or crime prevention, past data is increasingly used to predict future behavior. The growth of big data and cloud technologies is enabling these use cases. So what role does predictive analytics play in unified communications? What kind of use cases exists?

The basic requirement of course is that the interaction data from these channels are collected, aggregated, and mined to generate user behavior profiles. Based on these, the following use cases are possible.

Smart Sort

Every user has to browse through a number of offline communications at any time—emails, chats, and voicemails. These messages are presented to them in chronological order through independent applications. What if we create an application that combines all these interactions into one? And how about this application actually sorting these interactions in the priority you want to see them rather than in chronological order? For example, an email thread having a lot of threads in the last 30 minutes would get higher priority than an email sent to the entire organization. Machine learning can be used to understand past behavior of users based on which interactions he/she responds quickly to and use that for future sorting. This would help busy users pay attention to key issues especially when they are mobile or in a meeting.

Predicting Presence

One of the key features of UC platforms is presence. Presence helps platforms to route interactions to the user in an effective manner. The platform figures out the user’s presence through explicit settings, manual input, or device activity. But what if presence information is not available at any time? Can we predict presence? Based on the user’s past presence history and current information like calendar schedules, we can predict what the user is doing now and try to intelligently route interactions.

Predicting Intent

You see your desktop / phone ringing and the caller ID shows that it’s a key customer or your supervisor’s supervisor. You begin wondering what this phone call is for. What are they going to talk about? Do you have all relevant information at your fingertips? Wouldn’t it be great if there is an application that can predict the intent of the caller and pull up all relevant information on your desktop? For example, it can figure out that this customer is having a critical incident going on and pull up relevant records/emails. Past and current behavior can be used to predict future behavior. Predicting intent and then acting on that prediction to pull up relevant information would be of great help to all people, especially those working on the customer front.

Predicting Sentiment

Any kind of interaction can start going the wrong way. Emotions can start flaring and discussions can turn into arguments. It may be between peers or between customers and representatives. It is important for supervisors to sense that things are going wrong and intervene to set things on the right course. Real-time sentiment analysis on email threads or voice calls can help predict degrading sentiments and generate real time escalations to higher ups for immediate intervention. This kind of escalation will alleviate the need for supervisors to always stay “online” and watch for such incidents. Rather, they can focus on more important stuff and intervene only on alerts.

Analyzing Adoption

When organizations adopt UC platforms, their success depends on universal adoption, but not all users are going to get onboard from day one. They might continue to stick to their older communication means. In order to obtain true value from their UC investment, organizations need to promote adoption and take action when it is lacking. Machine learning can help understand user behavior and identify groups of users who need better training (or warnings). It can also spot trends of improving/declining user adoption. This information can help organizations to take corrective measures to ensure the success of their UC adoption.

Predictive analytics is in very early stages in the UC world. But it has a lot of value in creating better experiences for its users.

About the Author

As senior director of analytics and solutions engineering, Kumaran leads the efforts in delivering on data science and customer interaction advisor projects. His areas of expertise include performing data acquisition and cleansing, exploratory data analysis, visual data analysis, recommendations, modeling, scoring and simulation activities.