No matter where an organisation is in its analytics journey, all see the value in using more advanced techniques and strategies to better understand their customers by analysing the interactions with them more comprehensively.

We often use the “voice of the customer” via survey tools and other mechanisms to gain feedback on our products, services and processes. However, using the literal voice of the customer to genuinely understand sentiment, needs and frustrations can better help hyper personalise experiences, including those we are obligated to personalise from a regulatory perspective (think hardship, vulnerability, fit for purpose product or service).

It seems obvious that over time, organisations will need to adjust to be able to ingest and analyse customer interactions of all types to provide a better experience for their customers and agents (whilst reducing cost and risk).

When I started out in data analysis over 20 years ago, finding usable data was the tough part. Now the problem is the reverse. There are enormous amounts of data available – but finding it, and piping it to where it is needed and determining if you are allowed to use it for that purpose are just some of the hurdles to overcome when solving a business problem using data.

The same could be said of data related to customer interactions. Many organisations have a large amount of recorded conversations/unstructured customer data (voice, chat, message, email) and separately have clear value propositions where this data could be better used. In most cases strategy and roadmap bring these data sets together. They have invested in some great tools to record data and at times complete basic levels of analysis on interactions with their customers, but the potential for this data to be managed and analysed in a way that is efficient and effective has not been realised. They need to make the most of the money they have invested and have their data work harder for them.

Yes, this is an actual mathematically proven formula… OK, I’m stretching the truth a little there, but as a mathematician and seeing this concept proved in real life, I feel qualified and experienced enough to take some creative liberties. I have seen numerous examples of organisations seeing significant uplifts in efficiency and detection of compliance issues based on bringing together conversation analysis with customer data.

I’ve painted a picture of what this could look like in the opening paragraphs.

Sometimes our words say something, but our tone says something different. Many of us know this intuitively in conversation and can easily detect nuances in our own dialect or tone. Friends from overseas always mention they are confused with terms in Australia like, “yeah, yeah, yeah, nah”. Is it no, is it yes? Spoiler alert, it means “I understand your perspective, but I politely disagree with what you have just said.”. And we have all been told by someone at some point “yeah, I’m fine,” with a tone that indicates that clearly isn’t the case.

What is said is important, but how it is said can mean even more. Analysing words transcribed to text can be powerful but there is significant uplift in detecting accurate sentiment when using technology that mixes words with the acoustic features of a conversation including pauses, changes in tone, volume of speech and two voices talking at once.

What our customers say can tell us a lot. What our agents say, how they say it, how often they say it and what they don’t say… are all just as important.

We continue to see organisations starting to automate the detection of non-negotiables on calls such as completing parts of a process that are required from a regulatory or from a customer experience perspective. Voice monitoring technology can help organisations move from low sample analysis of compliance to much more comprehensive and statistically significant samples, using the same number of resources focused on the most likely/high risk examples of non-compliance. This not only improves efficiency but, in many cases, we have found that monitoring using certain types of technology can outperform monitoring completed by call assessors when it comes to the detection of certain risks.

Monitoring the agent component of our calls more comprehensively can also give us a better and more accurate read on agent quality and where training and coaching efforts should be focused.

Language and our use of it can change over time and is contextual. How an individual interacts with a provider will differ depending on a variety of factors, and their expectations will change over time depending on their tenure, the product, the environment, public perception, and personal circumstance. Being able to adjust models and strategies based on the individual, the cohort and even how language use changes over time will help organisations get the best out of their processes and strategies in voice and conversation monitoring.

The best version of this is where an organisation uses a process and technology that allows for immediate feedback on what is presented. For example, when a call has been incorrectly detected as a potential complaint, an agent has the ability to flag this as incorrect and the models driving the advice calibrates based on this feedback. A tool and process that allows you to provide feedback and learn from it, sometimes instantly, can make a real difference in the efficiency and effectiveness of voice and conversation monitoring.

So, there you have it, my top 6. If you are embarking on a voice or conversations journey or have some insights you would love to share, message me and I would love to have a conversation on this… only if it is recorded and monitored, of course.