What Is It?
Invoca’s Sentiment Analysis feature detects and displays positive and negative sentiment moments in call transcripts and provides an overall caller Sentiment Score. This data is also made available for use in call reports, dashboard tiles, and can power smart alerts (via Signals). Foreign language calls in Spanish or French are supported.
Who Has Access to the Sentiment Analysis Feature?
Customers with Signal AI Gold or Quality Management as part of their package will automatically see sentiment data added for all calls. No setup or configuration is required. This feature is available to customers in both the US and UK.
Sentiment Tells you Where to Look
Phone calls to businesses are one of the richest sources of actionable data available. Each conversation not only contains the caller’s intent, interest, and stage in the customer journey, it contains detailed insights on their pain points, preferences, and needs.
- For Marketers: Analyze sentiment trends by channel, campaign, web-page visited, product, or location to find critical issues affecting customer experience. Correlate sentiment scores to conversion data provided by Invoca Signals to uncover issues with pricing, availability, product issues, or how a promotion is being received.
- For Contact Centers: Identify agents handling more negative or positive calls than their peers to reveal coaching opportunities or provide relief to agents who may be handling a higher volume of challenging customer calls.
Accessing Sentiment Data
We have made sentiment data available in various places in the platform. The most noticeable place sentiment data is shown is when reviewing an individual call either from a report or from the Call Review Console. When looking at the details of a single call you will see the sentiment moments displayed alongside the transcript. Overall sentiment data for that call is displayed under the AI summary section along with the overall caller Sentiment score and the count of positive / negative moments for both caller and agent below this.
Use sentiment to filter calls from both the Call Review Console or calls reports to find calls of interest quickly.
How it works:
Call transcripts are analyzed and broken up into moments that are identified as either positive or negative. An overall Caller Sentiment Score is then calculated and provided as a data point along with the call being identified as positive (70-100) neutral (40-69) or negative (0-39).
Within the transcript positive 😊 and negative ☹️ moments are highlighted providing visual cues for both agent and caller throughout the transcript to help identify coachable opportunities or operational areas to address.
Note that the Sentiment Score reflects the sentiment expressed by the caller only, not the agent, during the conversation. It helps surface relevant calls based on how the caller felt during the call, but is not intended to assess agent handling or effectiveness.
Sentiment Score does not evaluate for heightened emotions specifically, but does take into consideration word count density. The higher the word count for a single exchange, the more weight it will have on the caller sentiment score calculation (either positive or negative).
For example, “word blobs” are a clear indication that a customer may be upset or frustrated so these exchanges are given more weight to help identify and surface more interesting calls for review.
It is important to note that sentiment analysis is based on the call transcription and does not evaluate the call based on tone and cannot recognize sarcasm.
Common Use cases:
Sentiment data makes for a powerful addition to any dashboard. Here are some common ways users are using sentiment to improve their marketing and call handling:
- Track Caller Sentiment by Marketing Source
- Track Caller Sentiment by Product or Service
- Retarget Frustrated Phone Leads to Win Them Back
- Create Lists of Extremely Negative Sentiment Calls to Be Contacted Directly
- Track Negative Caller Sentiment by Location or Contact Center Agent
One thing to note about sentiment
Sentiment is great at helping you find calls of interest, but we do not recommend basing agent feedback on sentiment analysis alone. Using call scorecard and evaluation form data is far more accurate for that purpose.
