Why is it difficult to understand customer emotions - and how can it be cracked?by
What is preventing a better understanding of customer emotions? A look at the challenges - and how they can be overcome.
As we explored in our earlier article, insights into customer emotions can be used in a variety of ways to improve interactions and the overall customer relationship.
However, the reality is that businesses are still unable to understand customer emotions at scale, and use this insight to predict what they might do next. And there are several obstacles that have historically prevented this.
Traditionally, organisations have collected data to represent / understand the 'Voice of the Customer' - everything from surveys to customer journey analysis. These are of course very useful for designing and monitoring the overall experience that you deliver (for example via Net Promoter Score and Customer Satisfaction, etc) and can be mixed with operational measures (such as Customer Effort Score or Average Handling Time).
However, Dorrington suggests that the data shows that these are very limited at predicting what the customer will do next. For example, NPS is a poor predictor of loyalty - even amongst 'Promoters'.
“You still need to do them to be operationally efficient and effective,” says Dorrington. “But they don’t tell you what the customer is going to do next. And your customers don’t care – so long as you are doing a good enough job, none of that stuff impinges on their consciousness. Whether you answer the phone in four rings or five rings is meaningless to them as long as it is within a reasonable time frame and within their expectations.”
Elsewhere, the rapid evolution of the likes of Machine Learning and AI techniques has meant that even though computers do not understand abstractions like emotions, there are techniques that can be used to ‘teach’ them about the indicators of emotions - things like facial expressions, voice tonality and the use of language.
These models are becoming ever more sophisticated, but once again there is a limitation hindering progress – while they are potentially good at giving you an emotional reading 'in the moment', emotions are fleeting.
Dorrington elaborates: “Emotions are difficult to measure (even if we are getting better at it) but also even if you can measure it, they are only accurate at that moment.
“Analysing voice or facial expression in real-time can give you a dominant emotion lead – this person is angry or happy for instance - but it doesn’t mean they were happy yesterday or will be tomorrow. Emotions are inherently unpredictable. And part of the reason they are unpredictable is that they are so dependent on each individual’s personal journey.
“The way which we interpret something that happens to us is very dependent on what happened to us in the past. That could be an individual encounter through to the way which emotions contribute to attitudes and beliefs and vice versa. Attitudes and beliefs tend to be more intrinsic and change more slowly but emotions flip and flop around all the time. So when it comes to operationalising it, even if I could tell you were upset an hour ago, that’s not necessarily how you’re feeling now. And because we don’t know what was triggering that, we don’t have the whole picture, so we can’t turn that into something like CRM.”
What is required to overcome these obstacles?
Dorrington proposes that in order for businesses to truly make the leap to understanding customer emotions and use this insight to shape customer interactions and relationships, they will need the following: methods of collecting data (especially 'unstructured' data), analytics tools to surface the emotions in play, matrix maths to fill in the blanks, and a method of operationalising the insights that represents the whole customer base (whether they have spoken to you recently or not).
Most organisations are already addressing the first part of this picture, in the form of customer analytics. However, Dorrington notes that analytical tools are only part of the picture, as you need data as narrative inputs and also context (e.g. customer journeys).
Analytical tools are only part of the picture, as you need data as narrative inputs and also context, for instance customer journeys.
He explains: “There are lots of sources of data but the one that is most important is to get as close to the authentic voice of the customer as possible. So that can come from voice when we change voice to text, or it can come from email or any other kind of chat or anything that gives you a narrative.”
The aim here it to extrapolate findings beyond the people that are being spoken with, to everybody else – to find out what the things are that influence the way people feel to get a running tally of how people feel based on an extension of the 12 core emotions, to find a way that they influence different groups of people without having to actually ask them how they feel.
Dorrington explains: “For example, if you get a penalty notice from your bank saying you’re unexpectedly overdrawn and you will be billed for £25 for being overdrawn, clearly it will make you feel negative. But I need to know how negative, because that can upset different people in different ways – to a millionaire a £25 fine is neither here nor here, but if you’re a single parent living on benefits it’s an enormous deal. So part of it is about how different people are affected by the event.”
The other thing that is important is how you already feel about the bank. If you already trust and like it then you’re more likely to forgive it than if you already dislike it. So both those things mean that the same event can be evaluated differently.
Dorrington says: “We can analyse the heck out of events and how they influence people but you have to know that there are different people responding differently and some of that is how they got to where they are now because that will colour how they see what has just happened. And it also affects the way that they anticipate what happens next.”
Other requirements for Dorrington’s approach include the ability to contextualise the customer feedback data – for instance, the likes of customer journey analysis.
He explains: “In banking, for instance, standing orders aren’t something that you even think about unless one of them doesn’t get paid and then it’s a big deal. Most stuff that happens transactionally doesn’t even impinge upon our consciousness, but when we see those things, some are more important than others and if we can identify what those are we can tell you how that adjusts the way that you currently think. So if you’re happy it might make you happier or less happy.”
Finally, Dorrington suggests that matrix maps are also required to convert the emotional analytics into a running score. Taken together, this potentially creates an impressive proposition: “Imagine a daily running score that gets adjusted for every customer, whether you’ve talked to them recently or not, taking into account not only the events and things you know have happened but also background things like the entropy effect.”
How to understand customer emotions
And it is precisely this that TTEC believes it can now deliver. It analyses customer narratives using Natural Language Processing techniques to surface emotions associated with encounters / touchpoints. To extrapolate this into longer-term predictions, it measures aspects of a fleeting 'moment', correlating it into part of a longitudinal journey using vector maths (a way of summarising a whole customer history into a few key numbers).
TTEC then worked out a way to then apply this beyond the initial customer group that is analysed to the whole customer populations, using data on what has happened in the past, how that's likely to be perceived and how it will impact future decisions. This emotion data is then appended to the customer record.
You are never going to be 100% accurate because you can’t know everything that happened to every customer today that is affecting their frame of mind.
“This is not a prediction of how the customer will feel in the future; instead, it is an explanation of what they are likely to do as a result of feeling that way,” notes Dorrington.
Finally, these insights are then embedded into operational systems – returning to our earlier example, removing a particular customer from a campaign because she is emotionally unreceptive.
Even with this approach, Dorrington does insist that organisations will have to set their expectations appropriately.
“When we’re talking about emotional analytics, you need to change your definition of success. It is currently about accuracy, but you are never going to be 100% accurate because you can’t know everything that happened to every customer today that is affecting their frame of mind. So if you think you’ll get a perfect read – you won’t. But it will be more accurate than anything else you’ve got.”
And Dorrington believes that we are nearing a watershed moment in customer emotion.
“Operationalising it, and being able to make decisions every single day, and adjust them on every interaction is the watershed moment,” he explains. “But what we have right now is something very usable and it breaks the cycle of ‘because emotions are fleeting they’re unpredictable, and because they’re unpredictable we can’t build a model’. Now we’re able to say that OK this isn’t 100% accurate, but we have a good enough read so that we can choose the path we take for the next interaction. We can understand how what we do affects individual customers based on their individual journeys.
“To do that you need modelling capabilities and machine learning, and the analytics, and you also need some market clever maths, and then you need the mechanisms to plug it into your operational systems. And then what you have got is a system that actually isn’t artificial emotions, but it understands emotion enough that it chooses a response that is better tailored to your emotional needs.”
And in a business world where organisations are looking for a point of differentiation, the ability to meet their customers’ functional AND emotional needs would be competitive advantage, for sure.
After two decades of experience working as a journalist and editor covering business and technology, including over 15 years as editor of MyCustomer, Neil now works as senior content manager at skills-based workforce management platform provider Spotted Zebra. ...