Four Steps - Using Data Analytics in Customer Service Solutions
We live in the world “powered" by the Web and communications where one in every three people have access to the internet and close to quarter of a million phones are shipped every hour globally. The amount of data shared every day is astonishing, reaching exabytes (millions of terabytes) every day. Analysis of the vast requires new techniques and tools that make our online experiences more personalised and effective.
Data analytics played a key role in the success of eCommerce in the last decade and has since manifested its value in a number of other domains including biotechnology, education or health care. The growth in volume, but in particular variety and velocity of relevant data led to the big data analytics trend.
However, in the customer service domain, the use of data analytics is far from optimal, leaving petabytes of valuable customer data unexplored and lowering the quality of customer service.
This can be attributed to many issues, with the lack of integration of various phases of the customer journey as the main problem. This makes the experience very frustrating and forces a customer to provide the same piece of information such as their account number multiple times in a short period of time. At the same time, a lot of valuable data about customer intentions or needs never reaches the agent answering the call.
These problems call for a better use of data analytics in typical customer service scenarios, leading to improvements in customer satisfaction and cost efficiency of the call center. Analytics allow for personalisation of the customer journey and can make the experience more tailored to the needs of individual users - from the moment they start browsing the site, to the second they finish call with a customer service agent. Here are the key requirements to make a better use of analytics in these scenarios:
- Collect data spanning the whole customer journey
- Apply analytics to better understand your customers and personalize the journey
- Use predictive analytics and machine learning to anticipate important events
- Continuously tune the analytics platform using feedback
There are a number of tools the techniques heavily utilised in data analytics that allow implementation of these requirements and significant improvements in customer service.
1. Collect data spanning the whole customer journey
Altocloud term: rich context data
The first step in harnessing the power of analytics relates to the availability of data about customers, they behaviour, profiles, and so on. Very often a single piece of information can make a big difference in converting a customer. This is particularly important in delivering effective customer service. Unfortunately, customer service agents lack important information - rich context data - when answering the customer’s call:
- What is the actual reason for the call?
- What does the customer see?
- What events in the customer journey resulted in the decision to call the agent?
- Have any of our customers faced a similar problem? How did we solve it?
Many of these questions can be addressed by analysis of the rich context data about a customer. Data that is available, easy and cheap to collect. No infrastructure is needed to start collecting it, in fact, if you are using google analytics you are already on the right track. All that is required is an extra line of code in your webpage, same way you do with google analytics, and you can start collecting valuable data about behaviour, location or needs of your customers as well as retrieve additional information their social network profiles to identify which other brands they like.
2. Apply analytics to better understand your customers and personalize the journey
Altocloud term: persona clusters and journey patterns
Analysis of the collected data is a key first step in understanding customers and their needs. Techniques such as segmentation (widely used in marketing) or clustering (one of the most basic yet powerful data mining approaches) can help to understand your customer and patterns in their behaviour. There are many tools and algorithms widely used for clustering that can be categorised in two groups:
- Supervised methods require examples drawn from existing data that are used to train the algorithm to categorize new data. For example, 100 examples of customers who did or did not buy the product will allow the system to predict if a new customer will convert or not. These approaches are useful when you have fairly good knowledge about the data, for example how many distinct customer segments exist.
- Unsupervised techniques identify patterns or structure in the data without any prior knowledge. For example, an unsupervised clustering technique will identify number of customer segments that should be formed and what are the characteristics of customers in each segment. These techniques are more complex to apply and customize for better results, yet prove very useful in many scenarios.
The most popular existing clustering techniques include Support Vector Machines, Neural Networks, Decision Trees and Bayesian techniques that have received a lot of attention in the recent years.
3. Use predictive analytics and machine learning to anticipate important events
Altocloud term: Action maps
Predictive analytics represent more advanced and sophisticated data analytics techniques. They extend the reach of insights into the future and provide powerful tools for predicting user behaviour, shift in user interests or future occurrence of particular events. In the customer service context, predictive analytics enable a number of powerful features and allow to:
- anticipate when a customer browsing the product website will need agent support
- identify and reach valuable prospects before they decide to call the agent
- predict the number of customers calling based on behaviour of your customers online
There are a number of predictive analytics techniques using statistical models that prove very useful. Analytics of rich user context data enables very interesting applications that help to make the customer journey successful but also lower the costs of running a call center. Predictive analytics engines typically exploit techniques based on regression models or Bayesian inferencing that achieve very good results in anticipating user actions, such as predicting that call will be initiated in the next 2 minutes or the spending of a new customer.
It is important to highlight that insights provided by predictive analytics platform are an important step in personalizing the customer journey, but require concrete actions that lead to measurable benefits. The real value comes from actions that decision makers, executives, or the marketing team take based on those insights. A good analytics platform will use sophisticated algorithms to create insights, a great analytics platform will provide concrete actions based on those insights.
4. Continuously tune the analytics platform using feedback
Altocloud term: Continuous learning
The behaviour and needs of customers can change rapidly similar to the assumptions about your customers and interactions making the carefully crafted models inaccurate. To cater for this, most data analytics techniques allow
The quality of predictions is achieved through continuous learning that incorporates both implicit and explicit feedback using real-time communications data:
- how your customers rate interactions with agents?
- how agents use the actionable recommendations
- which treatments can be applied to particular persona clusters