“You’re going to need a bigger boat”
Early in the film “Jaws”, the main characters go out on a fishing boat and have their first encounter with the great white shark. When they understand just how big and dangerous it really is, Brody, the Sheriff, then tells the captain: “You’re going to need a bigger boat”.
The volume of data that even a contact centre of 50 agents produces provokes the same reaction.
We have a contact centre with 50 agents. Each agent handles 40 calls a day. Each call has an Average Handling Time (AHT) of 8 minutes. When all agents are working, the contact centre produces 2,000 calls a day. If the average talk time is 6 minutes 30 seconds, then the contact centre produces 13,000 minutes of recorded calls per day.
Where contact centres have a manual listening programme, they listen to at most 2% of recorded calls. Even that represents 4 hrs and 20 minutes of recordings. How many 50 seat contact centres can spare someone to listen to calls for half a day, every day?
Where do we get our “bigger boat”?
The answer is speech analytics. However, the total cost of ownership (TCO) for an on-prem solution is beyond the budget of most 50 seat contact centres.
Using cloud technology’s “pay as you go model”, organizations can analyse the same data for a fraction of the price.
Amazon Connect and Omningage’s contact centre platform can be integrated with Amazon Web Services’ speech analytics platform, Contact Lens, to provide a wide range of speech analytics solutions.
What can we use speech analytics for? Classic use cases
Most organizations use Contact Lens, a speech-to-text solution, to conduct a full-text search on all interactions.
Sameem Smillie, Director of Global Contact Centre Solutions, says “The key to getting good value out of speech analytics often involves focusing on a single capability and then building a process around that feature.”
Speech analytics is commonly used to drive agents’ behavioral change and improve their performance or to drive process improvement.
This can be realized using keyword searches and various acoustic parameters, such as hesitations and sentiment to analyze what is being said, how it is being said, and by who.
Speech analytics can be used for a lot more than this.
Kun Qian and Magdalena Nedelcu describe how Contact Lens can be integrated with Amazon Connect Tasks to automate follow-up work, such as sending a customer an information brochure.
Our agent copies and pastes the customer contact details and the title of the brochure into an email for back office to handle. This might take her 90 seconds.
Contact Lens hears a phrase such as “I’ll send you a brochure”, then automatically generates the email request for the back office to handle.
Automating this task can save an agent 6 minutes per day. This is enough to handle another call. Productivity can increase 2.5%, from 40 calls a day to 41.
If you could build integrations to automate 3 more after-call work tasks, you could potentially improve productivity by up to 10%.
Improving CSAT and customer retention
Dominic Searle described how to build an automated notification system using Contact Lens, Amazon EventBridge and the Amazon Simple Notification System (Amazon SNS). It tells supervisors, in real-time, when agents were having issues with difficult customers.
Contact Lens monitors all calls for customer sentiment. When sentiment in a call falls below a certain level, a Contact Lens rule triggers EventBridge to create an event. This causes Amazon SNS to notify supervisors that an agent is having problems with a customer.
Here’s how this can save the business from losing customers and revenues.
Our 50-seat contact centre has an average agent tenure of 2 years. 2 agents are in training at any one time. Each customer’s business is worth £350.00 per year. Each agent handles 40 calls a day, of which 4 involve customers getting angry or expressing negative sentiments.
The training supervisor, on receiving the alert, steps in immediately, or silently listens to see how the agent handles the situation.
If our trainees have 4 problematic calls per day each, they handle 8 problematic calls where there is a risk of losing the customer, and therefore £2,800 of business per day. If the training supervisor prevents the loss of 1 customer per agent per day, she saves the company £700 per day. This adds up to nearly £200,000 per year.
There are 43 other trained agents working in the contact centre. They have the same number of “risky” calls but they can handle most of them.
Another rule can be triggered when the agent or customer mentions the manager or supervisor, and negative sentiment is detected. This happens 2 times per agent per week. It adds up to 86 incidents per week. If the manager can “save” half of them, she can save £15,000 of business per week, or nearly ¾ of a million per year.
What do the experts recommend?
We quoted Sameem Smillie’s recommendation to focus on one key capability. It’s important to have a clear idea of how you want to use the technology and what you expect to achieve.
Frank Sherlock recommends building a team to handle your analytics solution. It will need people to work on it regularly, and an organization with a culture of continuous improvement to get the best value out of it.
Ian Robertson recommends using speech analytics to divide calls into categories that are related to your business. It’s best if they are based on key phrases used by agents since they are more likely to use the same phrases than customers.
Experts recommend using speech analytics to help target manual call listening, but not to replace it. No speech analytics system can actually understand the calls and tell if agents are giving customers the correct information.
Experts also recommend not basing targets on speech analytics. This will drive your agents to say certain phrases in every call rather than try to improve the quality of their work.
Contact your technology partner of your Omningage Sales Director to find out more about how Omningage, Amazon Connect, and AWS’s speech analytics products can improve your contact center’s performance.
To find out more about Sentiment Analysis, read our recent post: