Machine learning models can be deployed on any data set that has labels you’d want to predict. Having the observations with the outcomes trains the model, which you can then apply to any new data set to accurately predict other outcomes.
It also helps fill in gaps. Customers are likely to take the time to respond to CSAT surveys with a numerical rating but are unlikely to provide an explanation in the text field unless they’ve had exceptionally good or bad experiences. Our platform analyzes entire call and chat transcripts and predicts CSAT scores and explanations for any customers who don’t participate. We do this by analyzing interactions with customers who also provided CSAT ratings, gathering key words associated with positive or negative sentiments, and using applied AI to compile customer insights. This gives you a more robust data set from which to analyze and gather insights.
Key Use Cases
Here’s how enterprises applied predictive intelligence to their customer analytics workflows:
Fortune 500 Financial Services Company
Since the company has been using our platform to analyze omni-channel data, we’ve been able to help fill in gaps by predicting customer satisfaction scores and reasoning for the 95% of its customer base who choose not to participate in surveys. These efforts have led to an increase in CSAT scores from the mid-80s to 92%, better trained agents who have effectively cut handle time by 20 percent, improved products and services (i.e. simplifying the credit card payment process), decreased calls and chats by building out a more robust FAQ section, and improved CX by making the website easier to navigate.
Fortune 50 Financial Services Company
This company’s efforts revolved around needing to create a holistic view of its clients’ experiences with various products by looking at omni-channel data. In order to uncover actionable insights, the company had to categorize customer data by complaint versus feedback to send to those respective teams. Within financial institutions, there are regulations about responding to complaints, so there’s a demand to detect complaints and respond as quickly as possible. Our analysts met with the company’s complaint management team and worked together to train the models to detect and categorize complaint versus feedback with existing data sets and deployed the machine learning models on new data sets. Our platform separated into the two categories in minutes, saving analysts weeks and months of time.
When diving into these customer reviews, the company found that several word pairings – such as “user friendly,” “international transactions,” and “interest rates” – all had unexpected sentiment attached. The company was able to dive deeper into the data to quickly find the reasoning behind the negative sentiment and use the information to improve products, services, and customer satisfaction.