Conversation Analytics extracts data out of calls, things like: agent empathy, sales performance, lead scoring, conversions, etc.
It does this by analyzing the content of the call using speech recognition technology and thousands of proprietary algorithms that automatically mine the call.
Last week we wrote a blog post briefly listing nearly 50 of the Conversation Analytics indicators already built into the system.
In this blog we’re going to discuss the Agent Empathy indicator
The indicator is used by contact centers, enterprises, and SMBs to determine if their phone agents are empathetic to customer and prospect concerns.
Empathetic employees can diffuse a situtation when an angry customer calls. The caller needs to know that the employee cars. Empathetic employees can actually increase customer retention substantially.
In short: agent empathy can actually improve revenue.
This indicator is expressed on a 0-100 scale. The higher the number, the more certain Conversation Analytics is that the event in question occurred on the call. For example, if the Agent Empathy indicator for a given call is 78, Conversation Analytics is very confident that the agent was empathetic on that call. On the other hand, if the Agent Empathy indicator was 14, Conversation Analytics is almost certain that the agent was not empathetic.
– Agent Empathy Alert – Conversation Analytics could send a Contact Center manager an email or an alert when the Agent Empathy indicator falls before 54 (or whatever). This allows the manager to, in near real-time, hold employees accountable for not showing empathy. And even call customers back when they have had a bad experience.
– Agent Empathy Call Back – When Agent Empathy falls below 70 (or whatever), a Webhook could immediately be sent to a CRM that places the caller into an immediate call back bin.
There are several hundred algorithms that Conversation Analytics runs on every call to determine each individual indicator. These consist of word combinations and phrases that help Conversation Analytics determine if the agent showed empathy on the call.
Keep in mind that Conversation Analytics also analyzes voice micro-tremors and acoustic cues (things like volume and rate of speech) to determine if something happened on the call. All of this data is viewed in context.
We’re not trying to figure out everything that was said on the call. In this case, we’re trying to answer the specific question: was the agent empathetic? This means that we can miss a few words here and there and still determine the gist of the call.
In this post we’ll discuss 32 different specific phrases Conversation Analytics looks for to figure out if the agent was empathetic. These represent less than 10% of the cues Conversation Analytics seeks for this specific indicator. We obviously don’t want to share every algorithm we use in a public blog post, but we also wanted to give you a behind-the-curtain view at Conversation Analytics.
Here are a few examples of what Conversation Analytics is trying to find for this indicator:
– Let me|let’s|let|I|we|we’ll|we’re|apologize
– I do understand
– I know what you mean|know what you|you’re saying|talking about
– I sympathize
– I understand situation|your point|where you’re coming from
– I wish I could help|do more|do something
– I’m sorry
– If I understand correctly
– Make sure I understand
– Must be difficult
– Sorry for the confusion
– Sorry for the delay|discrepancies|discrepancy|error|goof|screw up|inconvenience
As you can see, this gets extraordinarly complex. It requires extremely sophisticated algorithms, regression models and probit analysis to 1) find the phrases, and 2) calulate confidence and strength indications….and this is just for one indicator.
Again, keep in mind that the few phrases listed above represent only about 10% of the total phrases Conversation Analytics looks for within each call.