Features in these Releases:  Filtering Acknowledged Sentiments on Console | Smart filtering of Signals | Sentiment signal detection | ICA | Virtual Teams and Accounts | Custom Fields | Managing users

Filtering Acknowledged Sentiments on Console

Users can now filter out the sentiments that have been acknowledged already, and only focus on the ones that have not been acknowledged yet by anyone. This will help focus their attention on the cases that require their attention, and also keeps the page clean and responsive.

Smart Filtering of Signals

Not all Signals are created equally. We understand that and have now made it easy to only focus on the important signals all across our product. Our out of box configuration is smart enough and will automatically exclude some signals detected from outbound comments or internal notes. Additional customization is also possible if your organization would like to see more signals or more granularity.

Huge improvements to sentiment signal detection

Over the last few weeks, our ML team has been hard at work and made significant improvements to the engine, that has resulted in some incredible improvements to our signal detection. The results speak for themselves below:

• Churn Risk - 3600% increase in Churn Risk signals detected , indicating a high potential for customer churn based on their comments. This will allow users to proactively address any concerns and retain valuable customers.
  
  
• Lack of Progress - 1800% increase in the accuracy of detecting Lack of Progress signals, which indicates that a case is not progressing as expected and the customer has voiced their dissatisfaction. This will allow users to quickly address and rectify any issues.
  
  
• Positive Sentiment - 2500% increase in Positive sentiment signals detected, indicating that our customers are expressing a high level of satisfaction with our product.
  
  
• Call Request - 36% reduction in false positives for the Call Request signal, meaning that we are detecting more accurate signals when a customer requests a call. This will improve the response time and overall customer experience.
  
  
• Escalation Request - Improving accuracy from 82% to 96%, and reduced false positives by 64%. This signal is flagged when the customer has explicitly requested for an escalation. By increasing the accuracy, it’s now easy for the users to recognize and react to these requests.
  
  
• Confusion - 622% increase in detection of Confusion signal, which indicates that the customer was confused with the response provided. By flagging these more accurately, it now becomes possible for the users to course correct and provide better service to their customers.

ML improvements for Intelligent Case Assignment

Our Intelligent Case Assignment product just got some significant updates to its various machine learning models, resulting in more accurate relevant agent recommendations. The following models were updated in this release:

Agent-Customer History (Customer Experience)

We updated the Agent-Customer History model used to steer cases to engineers with a ‘good’ history with the customer for a case. This model was rewritten from the ground up for improved relevance.

Agent Availability (Backlog) 

* Improved ability to customize the model to a customer’s data and/or desired future behavior, in general

* Removal of underlying factors that made scores less interpretable

* Greater degree of model customization specific to how case Priority is considered by the Availability algorithm

* Greater degree of model customization specific to how case Status is considered by the Availability algorithm