Advanced Speech Analytics for Better Customer ExperienceSpeech Analytics is a tool used in every call center to track Key Performance Metrics (KPM). The use-cases are advancing day by day after the advancements in Artificial Intelligence and Deep Learning.
The data any contact center generates is incredibly huge via agent and customer call recordings every day. After mapping customer calls with CRM records. We can track and find patterns with everything that customers are saying like:
Keywords or key phrases that are trending.What percentage of customers who are calling in are not happy via sentiment scoring.Customers who are angry and might escalate.How does it work?
On the voice side, when a call comes in, there are servers listening to the voice of customers and agents. These calls get recorded as soon as the call gets over, which are sent out to servers to transcribe in English and then analytics is done on all the words or phrases used in the conversation between agent and customer. This is applicable for all the sessions via chat, email or calls.
This can be done for other languages as well.
What if (THINK!)
What if you know individual customer tone sentiment and average sentiment for an overall customer base?What if you gather data for each interaction and get the exact reason for all your customer issues based on their trending keywords and on the exact aspects they are talking about.What if you want to flag calls for Fraud/legal / bad language/sarcasm.What if you want to track agent sales and conversational skills by detecting misunderstanding, clarity of speech, pronunciation, speech tempo, too long/ too short pauses and other speech fetal parameters.What if you identify agent attentiveness via instances of mishearing, repetitions sought and questions missed.It’s not 100% correct but what it does is, it takes 10,000 calls which are received by your call center and drilled it down to 50 high alert calls which you really need to focus on because some issues you want to flag, are on these calls. Now, there is no need for hunting by Quality Analyst (QA), making them more efficient. We can flag calls in real-time also. By more advancement, it’s becoming better, like Amazon connect which provides real-time media streaming, helps in tracking calls instantly and flag in real-time.
It is much easier to run some type of analytics on the information delivered by the agent’s w.r.t disclosure and which customers need to know. It’s really hard to find out if an agent has read the piece of the script. This will ease to flag the calls if the agent didn’t deliver the full information. There are different ways to monitor what agents are not reading.
UseCases w.r.t Servicing:
1 — Improve QA process :
Today, Quality Analyst(QA) guys work more on listening and finding issue. Rather than making them police, it’s better to make them more efficient so that they can focus on training of the agents or the betterment of the agents by cutting there time on monitoring and focus on specific calls and agents.
2 — Track Customer Behaviour:
As to maintain the culture, it’s important to track customer attitude via sentiment score from each channel to give Customer Experience and Journey map.
3 — Sales:
We can generate more Marketing data and Customer data through trending keywords.
4 — Aspect based Sentiment:
We can provide sentiment scores for aspects like misselling, agent behavior, product, servicing, etc. which in result categorize calls.
Stop measuring 50% of the conversations:
In the past, we don’t have such analytics techniques to deliver such KPMs. Every time we think of what metrics make a world-class call center, so many times we measure wrong metrics which yield fewer benefits to deliver score-card on each call. We are taking handle time, FCR, TAT, etc. but we never measure the other side of the conversation which is the customer, we are only measuring those easy things that we can handle and control. But now with speech analytics, we can now track both sides of the conversation.
Finally, if we are measuring all kinds of metrics and have good sentiment scoring, that is the new standard for world-class call center.
Use Cases w.r.t Customer:
1 — Customer Satisfaction:
It really focuses on sentiment scoring, the issue with CSAT score or NPS is that you only get the two extremes, you only get the customers who are really mad or the customers who are very happy and you are missing all the details of the customer in between. Here sentiment scoring w r.t aspect like product coverage, details, invalid information, etc. helps looking into the details of each customer who felt ok, good or bad w.r.t each aspect. We can correlate negative sentiment from the customer to the negative keywords for eg if we have 30–40% of customers with negative sentiment, we can check those keywords or phrases like website issues, misselling, product issues, self-service model issue, etc.
2 — Agent Performance:
We can flag all agents interactions that are not up to the standards by categorizing calls by Agent Condescending or rude / Agent is dismissive / Agent is unknowledgeable.
3 — Understand Customer journey pain-points:
Customer journey mapping helps in the marketing side w.r.t at what stage the customer is complaining. Is he complaining about the IVR, Chatbots, product pricing, etc? If there is a pattern that becomes a trend then we can try to solve that by easing the customer experience journey. It helps in getting a pattern for solving marketing issues by keywords like too expensive, high fees, great products, etc.
Another benefit of speech analytics is to protect you from legal issues coming in through a customer by tracking phrases like IRDA, SEBI, BBP, lawyer, call my lawyer, etc. Any of those phrases or words have legal connotations, by detecting these we can send to the QA team or our clients to take actions. So there are many things depending on your business use case you can do like to detect fraud.
Real-life example — There is a company who is using this rough technology, an old Cisco PBX, they are using a real phone line, not on the cloud. No integrated chat, they have emails with different vendors as well, they have very little reporting. We seamlessly integrate it with their salesforce CRM, we provided them screen pops, emails, chat functionality nearly in less than a day all integrated together. They had 28 agents on phone, 14 on email, 6 on chat. So we reduce their staff to 12, which saves them tons of money, also gave them a better customer experience.
Here what we noticed though, we did all the awesome stuff, service levels are back to 80%. They didn’t even know what their service levels are before. When our client did customer satisfaction surveys, almost 40% of the customers are totally dissatisfied with the service so they come back to us for a solution , so we suggest them to turn on the analytics part of the software. We found out 55% of the customers have negative sentiment when they called us before actually talk to an Agent. So we looked at the negative keywords of why these customers are mad. We found never received my certificate, website issues, IVR kicked me out etc. so now they can take action on that. By getting, issues solved they have negative sentiment scores from 55% to 8–9% currently.
The power of speech analytics lies in providing you actionable data so you can take action to improve.
This might gives you a real view of what the world-class call center is.