It’s the bane of customers’ lives: waiting in hold music limbo only to be then greeted by an automated voice asking if they’re okay with the call being recorded.
Given the impatient state of mind of many customers, it’s probably unlikely many say no.
But what then happens to all of those recordings? Often, they’re expensively stored and never explored. But speech analytics can change that and bring valuable insight.
The idea of using call data for anything other than documentation, in many cases left to gather digital dust, can be daunting for businesses. Unlike sets of numbers, being able to extract and make use of voice data is not as easy as an excel spreadsheet. Its data gold is unorganised, or ‘dark’, and extra steps are necessary before you can make use of it.
Extracting and transcribing voice data, and then further analysing those structured call records for customer insights and business performance is the stock and trade of speech analytics.
Technologically, the process is pretty straightforward. A speech recognition program translates the voice data files into text, silence detection records where gaps in speech occur or identifies where people are talking over each, and emotion detection shows where, and how, feelings play into the call experience through changes in the customers’ speed, pitch and volume. Finally, natural language processing of the transcribed calls is used to identify the key themes, entities, products and sentiments so that further machine learning analysis can be applied.
At PwC, we take a multi-stage approach to where and how we use speech analytics, because to end up with insightful information, initial thought must be put into its application. Key questions a business is looking to answer need to be identified as the first step. Only then can the right calls be categorised and collected, and transcription and speech analytics be done in a meaningful way. That data is then augmented with additional customer and employee information, and advanced data analysis undertaken to gain actionable insight.
PwC’s multi-stage approach to speech analytics
There are many different insights that can be uncovered by analysing call data. While not all have an obvious cause and effect, enough data will indicate areas where more research should be done, or where quick tactical changes in the call centre environment will pay off.
Voice data can help improve sales effectiveness by identifying missed opportunities, which calls should be prioritised to specialists, triggers for escalation or adopting different approaches, and opportunities when customer service agents should push versus apply a light touch. For example, the data may reveal that a customer who mentions a product a higher than average number of times is willing to buy at the time of the call and could be encouraged to do so with the right personalised offer.
When it comes to losing customers, speech analytics can alert to when consumers are about to transfer to a competitor (for example, by how many times challenger brands are mentioned or how much emotion is shown), and identify opportunities and the right tactics to keep them. For instance, by passing them on to more experienced agents after certain prompts.
For customers themselves, experience with call centres can be a frustrating last resort. The ability to understand how a customer shows heightened emotion (for example, it may not be through outright aggression, but through confusion) allows a business to better train employees to respond, but also, to identify operational problems that are leading to bad customer experiences in the first place – before they even require a phone call.
As indicated by the automated message at the beginning of a call, being able to monitor and improve staff performance should also be an outcome of analytics. Data can indicate what good agents do and say, versus the actions of lower performing teams. Perhaps they leave more silences for the customer, put them on hold more/less or are subtler in their sales approach.
These are of course just some of the insights that can be gleaned in a product-based call centre, but naturally, there are many different areas that can be investigated for all kinds of calls.
From the above examples, it is easy to see that speech analytics can provide valuable insight, and an initial test case is often useful to gain broader buy-in across the business, but in our view, this is just the first step in the journey companies need to take to truly capitalise on their latent data assets.
For this to happen, the ‘toy must be turned into a solution’, with speech analytics embedded into the everyday, until it’s business as usual. While the test case – in defining a problem and following the approach outlined above – will prove the concept of utilising call data, it should then be embedded at scale.
To do this, organisations should work on rolling the speech analytics tools out across the business, embedding an action-research cycle as part of BAU processes and practices to drive real change. Leaders should be trained and encouraged to use the insights to make operational changes. Eventually, both the maturity of the analysis and the capability of the reporting to real time should be developed.
This might include building analytics into the broader customer experience approach, or real time analysis used to power ‘recommendations engines’ to give prompts and tips to agents in real time as they’re talking with customers. As AI and machine learning also grow in sophistication, new ways of working and data can be incorporated (for example, the use of behavioural economics, or social media data).
Adopting a mindset of continuous improvement will lead to better staffing, scripts, product offerings, competitor intelligence – and so much more.
With all of these benefits, businesses wouldn’t be blamed for thinking that voice analytics is a ‘silver bullet’ to better customer service. However, the reason many organisations have not embraced voice analytics yet lies in the difficulty of getting it right.
Translating speech to text is notoriously difficult. Accents, slang and sarcasm can all result in words and context not being translated properly, or tone being misinterpreted. Given the sensitivity of this data, many organisations are apprehensive about using cloud providers to help with the massive data processing required .
Additionally, linking a customer’s call centre interactions to their consumer demographics and purchasing behaviour can be difficult due to links to customer identification not being collected or kept up to date (for example, around emails, phone numbers and account numbers).
Recording call data should be more than a regulatory obligation or way of checking up on problems after the fact. Businesses should not be recording data just to prove that they can. Not only is it costly, there’s simply no point. Without an easy way to access the insights within it, it will likely sit in the cloud or on-premise taking up space and if it’s not adding any value, then it will eventually be deleted.
Making that call data accessible and applying analytics will open up endless opportunities for improvement, sales, customer experience and competitive advantage. By embedding the process and making any necessary behavioural or operational modifications, the insights gained from speech analytics have the power to drive change and add business value.
So the next time an automated voice asks you if your call can be recorded for customer service and quality purposes, say yes – and then ask them how.
© 2017 - 2024 PwC. All rights reserved. PwC refers to the PwC network and/or one or more of its member firms, each of which is a separate legal entity. Please see www.pwc.com/structure for further details. Liability limited by a scheme approved under Professional Standards Legislation.