Anticipating Customer Intent Better with Predictive Analytics

Anticipating Customer Intent Better with Predictive Analytics

How likely am I to cancel my bank account if I decrease the number of my routine transactions? Or what if I pronounced “I don’t like this” or “your competitor X does it the other way” to an agent during a call center conversation? Compared to an average customer, probably I would pose a higher risk of leaving this company. Maybe if the agent had the prior knowledge of me being risky, he/she would present a special promotion to me during the conversation and prevent this possible churn.

Considering that gaining a new customer requires much more effort and resource than keeping an existing customer, why take the risk of churn if there is a possibility to be alarmed before it happens?

Analyzing Customer Behavior

Collecting and analyzing past customer behavior is a need for sure. From small entities to large corporates, many organizations use these data to improve their services and enhance customer experience. However, it is not enough for today’s interaction dynamics. Companies have to be informed not only about the customer’s past actions but also about the customer’s possible behavior soon. It wouldn’t be wrong to say that being informed about the customers’ past actions is something; however, using this information to predict customers’ future behavior is a game-changer.

How Predictive Analytics Helps?

According to Markets and Markets, the Predictive Analytics market size is expected to grow from USD 4.5 Billion in 2017 to USD 12.4 Billion by 2022, at a Compound Annual Growth Rate (CAGR) of 22%. The main reason for this expected growth is being associated with the increase of the companies’ interests for forecasting future.

Predictive Analytics plays a critical role in predicting customer actions before they occur. It uses historical data to identify possible future behavior with the help of statistical algorithms and innovative machine learning techniques. For instance, historical data of customers that canceled their membership in the past would be a prior ingredient to train reference models for the “churn prediction” case. These models, which are trained for different scenario cases, can be used to make comparisons with the newcomer data.  Thus, it would be possible to see if customers matching with these patterns and to flag the interactions accordingly.

Predictive Analytics can also be used to detect fraudulent behavior before any serious damage is inflected. Companies can notice unusual activities in time that may cause incidents ranging from credit card fraud to fake identity calls.

Sestek Predictive Analytics

Besides typical historical data such as transaction data, demographic data, etc. the call center conversations also give valuable information about behavioral patterns. As an AI-based analytics company working on speech analytics, Sestek can create extensive prediction scenarios.

Sestek’s Speech Analytics solution currently analyzes acoustical and textual information of 1 out of 4 contact center calls in Turkey, which makes the case easier to expand this knowledge to get further insights about future customer intent.  Sestek records, transcribes, and analyses the calls between the customers and the call center agents by its own tools and combines these outcomes -like the speaker’s emotional behavior and tone- with further customer data to build reference prediction models for Sestek Predictive Analytics solution. These prediction models are trained for specific cases according to the organization needs. So far, our studies showed that Predictive Analytics can be beneficial for churn, fraud, and collection scenarios. It can be used to provide real-time agent guidance, next-best-action recommendations, or optimal marketing/sales offers for the benefit of both the customers and the organizations. Each new data bolsters up the machine learning algorithms and becomes more and more reliable compared to the instincts of agents.

Sestek Predictive Analytics is an ongoing project that has been financially funded by The Scientific and Technical Research Council of Turkey (TUBITAK). We “predict” that this new member of our analytics family will complement our solutions suite and present more comprehensive insights and actionable results for our customers.

Author: Tuba Arslan Kır, Sestek R&D Coordination Team Leader