Conversation analytics – a beginner’s guide
Forrester has identified emotion as having a bigger influence on customer loyalty than “effectiveness” or “ease” of dealing with a company. Ironically, companies have neglected programs to uncover customer emotion or sentiment, despite the technology being available.
Conversation analytics, also known as speech analytics, is slowly taking off in Australia as a way for companies to literally hear and measure their business phone calls en masse. After all, despite the drive towards digital, most B2B companies still prefer telephone over any other medium of contact. Until very recently, calls posed a mystery in comparison with channels such as social media, email or chat which, by virtue of the fact that the latter contain text, are more easily analysed. A desire to apply the same analytical rigour to the voice channel, as a source of customer insight, led to the advancement of conversation analytics.
What is conversation analytics?
Conversation analytics is a study of large volumes of phone calls between two or more people, whether recorded or in real time. It taps into a goldmine of insight – the conversations that go on every day between organisations and their customers, suppliers and employees – and constitutes one of the most reliable sources of authentic customer sentiment. The combination of language and audio processing used in conversation analytics offers an unprecedented insight into speakers’ emotion.
Speech processing methods
Conversation analytics draws on the solutions hiding behind two separate methods of speech processing. It combines them to create a full picture – what is being said, what is the language used, and what are the emotions behind utterances. The analytics process can be broken up into two stages:
Acoustic engine processing
This phase starts by interpreting the sounds of speech. It takes into an account the environment of the speaker, the telephony used to connect, the language spoken, accent and tonality. Ideally in this stage, there are about 4000 markers used to establish the properties of the individual words, including the human element, the emotion of the speaker.
Language processing gives a structure and meaning to the acoustic interaction. It uses episode-based language models that account for the vernacular used in a particular industry, location and situation, to create the most natural and precise maps of conversations. The higher the “tuning”, the higher the accuracy.
The key to a successful conversation analytics program lies in choosing the vendor best able to enrich the vocal stream and tune language.
Applications for speech analytics
Conversation analytics software is scalable and tuneable, meaning that analytics can be either applied to mine general data or resolve specific queries. Furthermore, the software can highlight recurring themes or focus on specific areas of conversations.
Enhance Customer Experience
In a literal sense, conversation analytics taps into the Voice of Customer. Data drawn from conversations is a direct account of customer needs, expectations and feelings, and a use-ready roadmap to improving customer experience.
Conversation analytics quickly pinpoints the shortcomings of scripts by spotting long pauses or negative emotions evoked by certain phrases or concepts.
By analysing every call, enterprise gains access to visible patterns. These patterns highlight the elements that call centre agents struggle with, or find hard to adhere to. They also enable the creation of personalised training tracks.
The root cause, keywords and topics can be cross-referenced with other metrics to extract insights that engage customers the most. Mentions of competitor names can be spotted and the context investigated – for example, a competitor may be offering a much cheaper price which customers are referencing consistently, and conversation analytics software can send alerts to appropriate stakeholders so they can take action.
Compliance and security
Conversation analytics enables monitoring of every conversation in real time, if required. This offers a unique opportunity to red-flag certain phrases and prevent data leaks.
Example: A health insurance company was attempting to reduce customer churn by offering them a reduced premium by eliminating coverage on services the customer didn’t. An analytics report discovered that their reaction to words like ‘reduce, ‘savings’ was negative, while ‘hospital’, ‘coverage’ had positive indicators. The company decided to change its script and started offering paid upgrades in the areas that evoked the most positive emotions. Churn went down, and an additional revenue source was created.
Conversation analytics is a technical answer to the Voice of Customer movement and a need to go beyond dry statistics. Its allure lies in its ability to recognise emotion and make those customer insights actionable. Its growing popularity is destined to increase with the advance of voice processing technologies. The irony and beauty of this technology is that it is helping companies understand the human dimension of their customer service.