Chat analytics: What every contact center leader must know
When you need to contact a company for support, would you rather pick up the phone or chat with an agent on the company’s website?
The answer might depend on the nature of your issue. While a slight majority of customers prefer picking up the phone to handle complex issues, chat is increasingly becoming the first choice for simpler issues. 70% of customers now use self-service channels like chat at some point in their resolution journey, according to Gartner.
If you operate a contact center, chat support should be part of your omnichannel strategy. And it’s essential to track the right chat analytics to ensure your customers and your contact center benefit from this self-service channel.
Chat analytics are metrics about the behaviors and language used in chat conversations between customers and a chatbot or live agent. Chat analytics fall into two categories: transactional data and conversation intelligence.
Transactional data gives you basic metrics about chat activity. For example, chat volume, interaction duration, and abandonment rate are all transactional metrics.
Conversation intelligence provides additional metrics that give you valuable insight into the content of chat conversations. This may include things like reasons for contact, causes of chatbot confusion, product mentions, and customer sentiment.
By combining transactional data with conversation intelligence, you can track trends in customer needs and issues so that you can provide a better overall customer experience. You can also identify opportunities to improve on your most important performance indicators, including improving your first contact resolution rate, reducing repeat contacts, increasing customer satisfaction, and reducing operational costs.
There are plenty of benefits to chat support for both customers and businesses. As a self-service channel, chat can be a low-effort way for customers to resolve their issues. In fact, when Creovai analyzed customer effort based on the language used in chat and voice interactions, we found there were half as many difficult chat conversations as phone conversations (7% vs. 15%).
In addition to giving customers an easy way to resolve their queries, chat support can also reduce a contact center’s call volume. Chat interactions cost less than voice interactions, so chat adoption helps reduce overall operating costs.
However, chat support doesn’t always live up to its promise as a low-effort channel. If chatbots or live agents frequently become confused or are unable to solve customers’ issues, those customers can become frustrated and may end up calling the contact center anyway. And every time a customer has to switch channels to resolve an issue, it has a dramatic impact on loyalty. Research from The Effortless Experience found that customers who pick up the phone after trying another channel first are 10% more disloyal than those who were able to resolve their issue in their first-choice channel.
So how do contact center leaders ensure their chat support is meeting their customers’ needs? That’s where chat analytics come in.
Chat analytics help contact center leaders understand what’s working well, what issues are causing the chatbot or live agent to become confused, and what’s causing customers to abandon chats—all of which helps them improve the chat experience.
The specific chat support metrics your contact center tracks will depend on your goals, but below, we’ve compiled a list of some of the most commonly used metrics.
Tracks the total number of chat interactions during a set period. This helps businesses understand peak chat times so they can plan staffing levels.
This tracks the number of back-and-forth exchanges between the customer and chatbot or live agent, which can help contact centers understand the level of difficulty to resolve issues over chat.
The percentage of customers who leave a chat before connecting with a live agent.
Also known as completion rate, this is the percentage of first chat interactions that resolved the customer issue (meaning the customer did not have to call the contact center or start another chat).
This refers to the percentage of interactions in which a rules-based chatbot fails to understand a customer query.
Also referred to as agent takeover rate, this metric looks at the percentage of chatbot interactions that were escalated to a human agent when the chatbot couldn’t meet the customers’ needs.
This scores the level of difficulty the customer experienced in resolving their issue over chat. While some contact centers rely on survey responses for this metric, Creovai predicts an effort score for every conversation based on the words and phrases the customer used.
A score indicating how satisfied the customer was with their chat experience. As with customer effort, many contact centers depend on survey responses to track this metric. However, Creovai uses AI to predict how every customer would have rated their satisfaction after an interaction, regardless of whether they completed a survey.
While most chat support software reports on basic metrics, such as volume and duration, few offer deeper insights into where the chat experience is going wrong. To truly understand the chat experience—and how to improve it—contact centers need conversation intelligence software. This software analyzes what agents and customers say and how it’s impacting core metrics, such as first contact resolution and CSAT.
Below are several types of chat analytics that conversation intelligence can help contact centers track.
Predictive analytics come from using historical data, statistics, and modeling techniques to predict future outcomes. For example, Creovai uses AI and machine learning to predict how every customer that chats with a business would rate their satisfaction and effort. Our platform also analyzes chat conversations for leading indicators of churn risk, helping businesses know when to proactively follow up with customers.
These types of predictive analytics help businesses understand how customers feel about the chat experience without having to rely on post-interaction surveys, which typically have low response rates. They can also help businesses uncover why customers feel the way they do so the business can take direct action. For example, if the business found a correlation between chat conversations about their warranty policy and customer dissatisfaction, they might reevaluate their policy or educate agents on better ways to communicate about it.
It’s not enough to know the percentage of chats that included a fallback response or escalation. Contact center leaders need to understand what topics are causing their chatbots or live chat agents to become confused.
Conversation intelligence software like Creovai can identify contact reasons, topics, and even specific phrases associated with high rates of chatbot or agent confusion. This allows contact center leaders to train their chatbot or live agents to address these issues. Contact center leaders can also use Creovai to analyze how their most successful agents are resolving these issues in voice interactions, giving them new scripts to use in their chat support.
Why are customers abandoning chat interactions before getting their issues resolved? In some cases, it may come down to not getting a fast enough response from a live agent. But in other cases, it’s more complex and can be a sign that a chatbot or live agent is failing to deliver what the customer needs.
As with chatbot and agent confusion, conversation intelligence software can analyze interactions to identify trends in chat abandonment. This can help contact center leaders identify parts of scripts or workflows that are creating friction so they can make improvements.
Are customers going to another channel when they can’t resolve their query over chat? One way conversation intelligence can track this is by flagging mentions of channel switching in voice interactions. For example, if a customer said, “I asked your chatbot” or “I was chatting with another agent,” the software would label this as a channel-switching instance. This helps contact centers identify the contact reasons or topics that are most likely to lead to a chat escalation. Armed with these insights, they can make improvements to their chat scripts or agent coaching—or streamline the process of directing customers with more complex queries to their voice channel.
Chat analytics are only valuable if you put them to work. Here are four examples of strategic ways to apply chatbot and live chat analytics in the contact center.
Your voice interactions contain valuable insights into what your top agents are doing well. By analyzing your best voice interactions, you can see how your most successful agents are approaching customer issues, down to the specific phrases they use. You can use these interactions to train your chatbot, whether that means adding the interaction transcripts into your training data for a GenAI model or building the best responses into your workflows for a rules-based chatbot.
By isolating your chatbot interactions that include a fallback response, you can identify the most frequently asked questions that cause chatbot confusion. This will help you prioritize which new responses you need to build into the chatbot’s workflow. This is another opportunity to model your responses off your best voice interactions. (One of our customers, TwinStar Credit Union, took this approach and reduced their fallback rate by 75%.)
Conversation intelligence software can identify phrases associated with live agent confusion (e.g., “I don’t know,” “Let me check”) so that managers can see which agents are struggling with which topics. This enables managers to tailor their coaching to help their agents resolve issues more efficiently over chat.
Any time a customer abandons a chat or escalates to a live agent, they’re experiencing effort, which increases the likelihood of churn. Chat analytics can help you get to the root causes of effort so you can address them. By prioritizing the most impactful improvements to your chatbot or live agent training, you can improve your first contact resolution and increase customer satisfaction with your self-service options.
If you offer customer service across multiple channels, you need to create a consistent experience across those channels—and enable customers to resolve their queries in their first-choice channel whenever possible. Chat analytics help you ensure your customers get what they need when they choose self-service, creating a low-effort experience and deflecting calls to your voice channel.