CX Analytics: Taking the Complexity Out of It

The problems with customer experience initiatives  

Very often executives focus on the newest data or the latest modelling techniques they need to implement in order to boost accuracy by a few percentages. However, to make an analytics project successful, there is a more important factor to consider.  

Most large businesses have more than enough customer data and don’t need anything ground-breaking in order to boost business performance. What executives need to focus on is the end goal of a customer analytics project. The business goal they are trying to achieve and the expected return on investment should be at the centre of their strategy. Not surprisingly, that has a lot to do with simplifying customer experience. 

In the field of customer experience analytics, sadly, the end goal in 4 out of 5 projects is a ‘high-level improvement in experience”. The expected ROI is often unclear and thus passed as strategic. There are three reasons for this problem:  

• Existing data usually does not cover the full customer base because response rates are very low. In most B2C industries response rates to satisfaction surveys vary between 5% and 15%.   
• It is considered the domain of customer experience executives and is rarely used in combination with other types of data and as part of cross-team initiatives. 
• Often the relationship between customer level satisfaction and lifetime value is overlooked, which hinders the potential returns of improvement initiatives  

There are no average customers 

CX analytics looks at an average customer when, in fact, there are no average customers. As a result, customer experience initiatives tend to only look at the average results to create company-level or touchpoint-level initiatives. Because they are designed for an average customer, they have an average impact.  How to evolve beyond that? By using existing data more strategically, in a way that simplifies decision-making and allows for scalability. Instead of solving one average customer issue today, the CX team can be solving a thousand personal and urgent customer issues with an immediate impact.  

Simplifying customer experience at scale and focusing on impact 

The number one priority today for both a CX manager and an analytics leader should be to simplify decision-making at scale and focus on customer-level impact. The big question is how to do it. The experience of a telecommunications customer, for example, is comprised of multiple different factors. Even if you take one key driver for satisfaction – connectivity – there are a lot of other external and internal factors at play. There are a lot of things that could go wrong at all times and a lot of them aren’t immediately fixable.  

There is no need to gather additional data or overcomplicate models in order to become an impact leader in your organisation. Instead, help your organisation prioritise the one thing that will be most impactful – for the customer experience and for the bottom line of the business.  

How to identify that one thing you need to prioritise? Connect the dots. The gaps between the dots are all those challenges we have looked at.  

• Connect the dots from customer to customer. From the ones that respond to surveys to the ones that don’t.  
• Connect the dots from customer needs to actual improvement.  
• Connect the dots from the initiative to the business impact  

Once you connect the dots you will enable personalised, customer-level service at scale and will be able to measure the actual immediate and long-term impact of CX initiatives on financials and business growth. 

Divide and conquer  

Before we get ahead of ourselves, let’s explore the connection of the dots in more detail and understand how it works in practice.  

How to fill in the gaps from customer to customer with predictive NPS  

One of the big industry challenges, other than simplifying customer experience, is knowing how satisfied your customers are – all of them at any moment in time. We solve this challenge (for our clients) with a model called Predictive NPS. Predictive NPS accurately predicts satisfaction scores for the whole of your customer base at any moment in time, not just after a key interaction when you send a touchpoint survey.​   

In this model, we combine behavioural data (what products customers use, do they contact the call centre, etc.) with relational NPS data from customers who have already responded to traditional surveys. 

The algorithm then identifies which factors in customer behaviour have the highest impact on the relational NPS of a specific customer. Once you are able to predict likely detractors, you can take proactive measures to improve their experience with the company and most importantly – prevent some of them from churning. And the best part is that you can directly measure the impact in terms of saved customers and retained revenue.  

Here is an example of how this works in practice: Imagine John – a mobile operator’s client, who pays his bills on time and based on his transactional data is perceived as a loyal customer. But because John is a frequent business traveller, every now and then he gets shock invoices and is unable to resolve his problem by calling customer service. He is very unhappy but doesn’t fill in any surveys and the company doesn’t suspect he is a possible churner until they run Predictive NPS and actually find out he is a SuperDetractor. A single call with a personalised offer for him allows the company to retain 2000 EUR annually in revenue from John and most importantly – gain a loyal customer.

How to connect the dots from customer needs to actual improvements  

Making sense of customer behaviour, especially if you have tons of customer feedback (from open text questions, social media, call centre, etc.) can be hard, but text analytics and advanced analytics models will come to your aid.  

Starting with NLP – driven text analytics models you can make sense of this open-ended feedback, classify it into topics, track positive and negative sentiment and link it to explicit satisfaction scores given by the customers. Then, root cause analysis helps get to the bottom of things that often remain at a very high level. For example, a lot of broadband Internet subscribers may share that the plan prices are too high. But when you dig deeper it turns out that price is relative to Internet reliability and speed. If you can’t deliver on your promise for a quick and reliable connection, then the fees you are asking for indeed seem too high for the customer.  

Classify topics into three categories: hygienic, must-have and nice-to-have factors by using advanced KANO analysis and prioritise the topics that will bring the highest improvements to customer satisfaction. Financial linkage models track how much potential revenue is lost every time there is an NPS dip.  

Finally, a customer won’t always share directly the true reason for their dissatisfaction. Linking CX drivers to operational KPIs helps to identify any underlying reasons for dissatisfaction. It also allows us to better understand customer expectations, align internal KPIs accordingly and to simplify the customer experience programme. 

Connect the dots from improvements to the expected results   

Understanding how customer lifetime value changes in real-time allows to improve the time to react to high-risk events and clearly measure the impact of any rescue efforts. The main drawback of standard CLV models is that they have a fixed flag upon which they calculate propensity to churn (6 or 12 months).   

In contrast, the Dynamic Customer Lifetime Value model calculates customer lifetime value for any chosen future moment in time which allows for greater accuracy and better planning.  

The model has 3 main elements:  

• Churn model – predicts the probability to churn for each client in each and any future point in time  
• Cost-to-serve model – predicts the total cost-to-serve for each client  
• Revenue model – predicts revenue for each customer which together with the probability to churn and the cost to serve gives an estimate of the customer lifetime value  

Imagine the following scenario:  

• A customer is experiencing connectivity problems  
• They call customer service and continue to visit support pages afterwards. Their behaviour suggests that they are a likely Detractor with an unsolved problem, who is likely to churn  
• Their likelihood to churn increases, but at the same time it’s a valuable customer with high predicted CLV and you want to retain them  
• Customer agents are notified and contact the customer back, solving the problem  
• As a result, the predicted CLV is even higher than it was initially – you now have a more loyal customer than before  
• At the end of each day, you know the exact number of retained revenue   

The Impact on customer lifetime value   

Employing the right predictive models for successful prioritisation has an impact on the three main elements of customer lifetime value.  

• Decreases cost-to-serve – getting in touch proactively with customers with low predicted NPS scores leads to a noticeable reduction of care interactions.  
• Helps to retain or even increase the expected revenue from the treated customers.   
• Impacts acquisition costs for new customers. The most straightforward way to do this, is to use predicted NPS scores to activate referrals from previously unsuspected happy customers   

An award-winning success story  

We had the chance to do the exact same project with a major German telecom. To solve the problem with proactive and personalised customer service at scale, The Customer Event Account was created – a client dossier that allows customer service agents and other people from the company to understand their customers better and act accordingly.  

It had two main elements – predictive NPS  and The Next Best Action module (a powerful suggestion tool). How does it work? While on the phone with a customer, call centre agents can open it and check for the best possible reaction to the customer request. They are free to use those suggestions in the conversations or decline them if they have other considerations  

Together Predictive NPS and NBA make up the Customer Event Account and become a powerful tool to provide proactive and personalised service. Our client is now able to help customers proactively. To date, over 16 000 customers who were suspected to be experiencing technical or other types of issues have been contacted proactively by phone or e-mail.   

Those customers also receive more relevant upsell/cross-sell offers. How does that work in practice? Customers who are predicted to be unhappy may be left out of certain upsell campaigns because we know that they are very unlikely to take the offer.  

The Customer Event Account experts gain a better understanding of the most important repetitive issues on a company level and are able to show a significant impact on business results. Since the start of the project, our client has been able to retain 13% more Internet customers and 17% more cable TV subscribers that otherwise would have churned.  

Where to begin? 

If you are already entertaining the possibility to unlock a tangible impact on revenue, acquisition costs and cost-to-serve, you might be wondering where to start and if this will be difficult or time-consuming. 

Start with what you already have – the Voice of Customer program. Scale your knowledge of the customer with Predictive NPS and then – scale the impact by successful prioritisation. We can make sure the process is neither difficult nor time-consuming.