Speech analytics: the latest contact centre conversation


Share on LinkedIn

Anyone who’s ever made contact with a call centre will be familiar with the following phrase, or some variation of it: “Calls may be recorded for training and monitoring purposes.”

It’s one of those sayings that’s become so commonplace, few would bother to gaze beyond its surface-level meaning.

It’s not difficult to imagine how bad calls or good calls could be used to instruct new employees, or even provide evidence in a dispute further down the line. And that’s typically about as much as call recordings are used for.

But all of this is changing.

Developments in speech analytics technology are opening up a whole raft of new possibilities for contact centre managers who want to improve performance and customer satisfaction rates.

What can speech analytics tell us?

Put simply, better speech analytics allow contact centre managers to extract more data from call recordings.

This is something that’s been toyed with for a long time, but until recently the technology either hasn’t been sharp enough, or contact centres haven’t had a sturdy enough tech stack to deal with it.

Developments in speech recognition technology and the steady trend for contact centres to adopt more flexible and scalable cloud solutions has created an inflection point for better data analytics.

Speech analysis offers a whole range of data outputs. There are macro measurements of things such as dead time or crosstalk, as well as more granular observations: customer sentiment, for instance, or identifying transactional requests (versus lengthier more nuanced discussions).

Some of these data points will have obvious, immediate uses. Negative sentiment detected on a call might require follow-ups, too much dead time could indicate script holes or slow on-screen prompts on the agent side.

Crosstalk could be ironed out with further training, frequently occurring transactional requests could be dealt with via interactive voice response (IVR) to free up agent time.

On top of this, we can go further and think about how voice integrates with other self-service elements, such as chatbots or knowing when to deflect the enquiry to a part of the website. It’s all part of the omnichannel communications mix.

What’s most exciting about this technology, though, is its possible future implementations.

What will speech analytics be used for in the future?

As well as using analytics to identify requests that can be handled via keypad IVR, call recordings can be used to train voicebots to offer a more efficient (and natural) customer experience.

Beyond this, running keyword analytics can be used to identify unforeseen or undiscovered problems – a water company receiving numerous calls in a short space of time with the word “leak”, for example, or a smartphone manufacturer that keeps hearing the word “battery” and “fire”.

The same approach can be applied to develop new offerings: mining customer queries to identify patterns in how a product or service might be improved.

Bi-directional streaming, such as that offered by Twilio Media Streams, can be combined with an analytics platform to provide real-time data on live calls, automatically identifying moments, for example, when a call may need to be escalated or a manager may need to step in to protect an agent.

A similar setup could be used by charity helplines and emergency responders too, for whom speed and accuracy can make all the difference.

Speech analytics can also be used to offer an added layer of protection for callers and customers: Santander has been testing the use of voice recognition as a form of biometric security since 2019.

“These analytics may be used for training purposes”

It should be stated here that much of this technology is still in its early stages, and, as with any new solution, it’s worth taking a gradual approach – to begin with, at least.

If your business hasn’t dipped much more than a toe into the world of data analytics, then voice and speech analysis might be a tricky place to start.

Instead, consider developing your analytics engine using webchat systems and chatbots, where the language being used is clearer to interpret and the outcomes more immediately measurable.

Over time, you’ll be able to add more functions and features – until eventually the readout at the beginning of each call says: “Calls may be recorded for training, monitoring, and customer service improvement purposes.”

Jordan Edmunds
I am business development manager at Zing, a dedicated Twilio consulting partner. My role is to help organisations put their contact centre at the heart of customer engagement. I'm involved in the whole process, from speaking to new customers to assess their needs through to the full implementation of Twilio.


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here