Do you know what your customers really think about you?

Research carried out by the Cambridge Service Alliance has shown that widely used methods for gauging customer satisfaction – such as ratings surveys and Net Promotor Scores – do not really tell you what your customers are thinking. Dr Mohamed Zaki and Professor Janet McColl-Kennedy, worked with CSA partners to devise a better way of understanding the customer experience.

 

According to a 2018 analysis by PwC, customers are so used to great customer experience that 32% say that they will walk away from a brand after just one bad experience. This is a sobering finding, particularly if you have no idea when your customers are having those bad experiences. This will often be the case for those firms that rely on surveys to score their customers’ satisfaction levels.

 

Our research has shown that high overall satisfaction scores do not mean that your customers are happy with you. We found that while a large percentage of customers gave scores of 8.5 out of 10 or more, 90 per cent of them used the comments section to voice significant complaints. But their complaints were not being addressed because the high scores were lulling the firm into a false sense of security. Further analysis revealed that this lack of response resulted in lost sales. For instance, one so called “satisfied” customer reduced purchases from over $200,000 to less than $2000.

 

The challenge: why surveys don’t work

Customer loyalty is complicated and, as PwC points out, it can be fragile – and surveys are a blunt instrument. Firstly, only a small subset of customers ever complete them. One CSA partner, reported only 1-2 per cent response rates. Numbers like that are never going to give you robust results. Even if all of those customers were saying the same thing, it is easy for senior management to ignore them, on the basis that the sample size is not statistically significant.

 

Secondly, surveys can only ever give you a snapshot of a moment in time. How will you know, for example, about a customer’s bad experience if it takes place the day after you last surveyed them? And timing is another challenge. Firms often send out their surveys weeks after the last touch point, by which time they are not capturing real feedback as the customer’s emotional response will have well and truly dissipated and key details of the interaction are likely to have been forgotten.

 

Another critical weakness of standard ratings surveys is that they cannot pick up those emotional responses in the first place. Turning customer feedback into one-dimensional metrics has been a rational organisational response to the challenge of interpreting (and, hence, being able to act on) large volumes of data. But this is not helpful if those headline numbers are preventing the firm from hearing what their customers are trying to tell them.


Finally, there is a danger that carrying out surveys can make things worse rather than better. If you ask for a customer’s insights, you need to be able to act on them. If you don’t, you end up making them even more frustrated.

 

Building a better solution

Our mission was to design a prototype that enables firms to overcome the limitations of conventional survey techniques to capture meaningful feedback that can be used to intervene when things go wrong at an individual level and to develop new services at a systemic level. By using cloud and AI technologies we wanted to be able to track the customer experience in real time and create a user interface which would allow service providers to see what’s going on and be able to act on it.


We started by mapping key touch points and then developing methods of collecting feedback that customers would not find intrusive or irritating. Surveys remainan important weapon in the customer experience armoury, but we also wanted our system to use every contact between the customer and the firm – such as emails and phone calls – as an opportunity to extract useful insights.

 

To make sure we got our approach right, we worked with the customers themselves. Not only did this give us a better outcome, it also meant they were invested in the process and more likely to make it work.


For our prototype, we identified three critical touch points and embedded feedback mechanisms in each of them. The first is when a call centre sends an email to a customer confirming a job has been booked and includes a link to a survey. The second is when the technician is on site and they ask the customer to complete a survey on a tablet there and then or they email it to them later if preferred. The third and final touch point is when the invoice is sent, accompanied by a link to a survey.

 

Getting the survey questions right is also key. The trick is to keep them simple to make it easy for the customer to complete. But the real game-changer here is Machine Learning. Until now it has been very difficult to interpret and act on qualitative data collected in large volumes, hence the tendency to reduce it down to numbers.


Machine Learning allows us to combine conventional rating scales with analysis of the customers’ views – written in their own words – to get a real insight into their emotions: are they feeling love, joy, anger, fear or surprise? It also makes it possible to ask customers to suggest ways in which the services can be improved – and be able to turn those insights into meaningful outputs.

 

Usability is as important for the service providers as it is for the customers. For the former, a dashboard can show in real-time how particular areas of the business are performing. If there are issues emerging it allows users to drill down to see what’s going on and intervene. It also gives users a view across the entire customer journey, exposing problem areas. And it means that different people across the organisation will have the same view of the customer, so that if problems have arisen in the past all customer-facing staff will be able to see what has happened and act accordingly.

 

Where next?

The CSA team built the prototype using four years’ worth of real customer feedback as the training data for its Machine Learning. It has now been used with customers and successfully demonstrated the proof of concept.


Potential next steps will be to integrate it with existing company systems to further enrich the view of the customer and enable at-a-glance analysis of customer experience related to other metrics such as customer spend. We also envisage that
advances in technology will allow us, for example, to use voice recognition technology to capture emotions expressed on phone calls which can then be analysed both for content and tone.


To design and deliver successful services you need to know what your customers are thinking, feeling and doing. Until now, this has been difficult to find out but Machine Learning and new data analytical tools have fundamentally changed the rules of the game. There is no excuse now for not hearing what your customers are telling you.

 

 


 

Find out more about the work carried by the Cambridge Service Alliance, including customer service analytics, at the upcoming CSA Industry Day Conference, 16 October 2019