Know more before you act:
The role of big data in the
customer experience
By Neil Lilley, November 5, 2014
Improving customer experience is the top priority of every communications service provider, and for good reason. In a world of brisk innovation, customer expectations of service variety, availability, quality and ease of use are rising higher and faster than ever. Operational excellence is obviously key to meeting those expectations, but first things first. Before they can successfully deliver an excellent customer experience, operators must first answer two questions: what is the customer experience they are currently providing, and what experience do their customers want? In other words, operators must know more before they act.
In a way, providers already have the answers to these questions. It’s contained in the data they’ve been amassing every day, on every customer interaction, every order, every trouble, every event and every customer usage session. There has been simply too much data, accumulating too fast, for anyone to make sense of the detailed insights hidden within it. Now, however, big data analytics are changing everything. Combined with first-hand customer research and deep network and
operational knowledge, they can give operators new insights into everything from customer behavior and preferences to the root causes of the network events impacting the customer experience.
Adopting a structured approach across business process areas
Every key process of service provider operations has an impact on customer experience – idea-to-implementation, plan-to-provision, lead-to-service, service-to-cash, and experience-to-resolution. A structured approach to answering our two questions for each key process will provide the most useful information, and also the flexibility to extend that approach in the future.
The first step is to determine what metrics can measure the quality or success of each of these key processes. This will often mean correlating data from different and diverse sources. The second step is to assess what improvements should be made based on the data-driven insights. The third step is to implement or refine capabilities such as network planning, customer care or quality monitoring in order to execute against the new goals. Finally, new measurements are taken to provide a feedback loop and continuous improvement.
For example, consider the idea-to-implementation process, which determines whether the operator has created the services and bundles that the customer wants. Operators have a wealth of data on how customers use services, including when, where and what content is consumed; how those patterns correlate to time of day, day of week or type of location; services or apps frequently used together; usage volumes; and CRM and other data. The volume of this data, especially usage data, can be overwhelming, but with a big data architecture these volumes can be handled. As a result, operators can assemble user profiles for individual or aggregate customers. One example profile might be for users who consume news during commuting hours, conduct video chats during midday and upload photos on the weekend. Operators can then develop tailored service packages, including devices, pricing, bandwidth, apps and other components that will attract and retain such customers.
Improving The Experience
Identify Desired Experience Measu r e & A na ze tE ely h xperenc i e I mprove & O timize thep E xperenci e MEASUREMEN T INSIGH T DECISION SUPPOR T AUTOMATION
Big data brings it all together
Another very important example is the experience-to-resolution process. It comes into play every time the customer uses a service, and it has the leading role in monitoring and improving the usage experience. Trying to monitor customers’ usage experience in recent years, carriers have relied on a combination of service quality management (SQM) and probe solutions. Each is valuable, but each also has its limitations. Probes, for instance, can measure how the customer might be perceiving quality by monitoring lag, jitter and other quality metrics. However, probes have no insight into the cause of quality issues in the network, or elsewhere. They cannot, for instance, differentiate whether frozen streaming video is caused by device misconfiguration, network congestion or content server overload. SQM tools, meanwhile, can provide excellent root cause information, but they do not provide the individual customer experience view needed by customer care.
By implementing a big data solution for the experience-to-resolution process, it is possible to unify these two types of data. Done properly, individual user session records can be correlated and analyzed in real time, for an unprecedented level of understanding of the quality of a customer’s experience as well as the network or other events that cause it. Probe-like data can be used to determine the key performance indicators (KPIs) that tell whether the customer’s experience is satisfactory. When these KPIs fall beyond a defined threshold, the exact, time-correlated network events that led to trouble can be identified.
The power of this approach cannot be overestimated. Imagine being able to see that a customer has been denied access to video content, and at the same time immediately know – based on the actual, correlated network events – whether that denial was due to device errors, RAN congestion, traffic overload at a particular node, mismatched protocols, content provider downtime or other causes. Empowered by this approach, customer care can provide a much more rapid and satisfying answer to customers’ inquiries, while greatly reducing call times and volumes. Likewise, the service operations center (SOC) can see these problems as they happen and rapidly identify the cause, resulting in reduced mean time to repair (MTTR) and, therefore, a reduced number of customers who will encounter the particular trouble.
Combining the right resources
Several resources are necessary to successfully apply these approaches to customer experience management. A big data
platform has to provide the capability of handling vast amounts of real-time data. But the real value is created by the expertise that determines which data elements, in which combinations, at what thresholds and from which diverse sources are needed to derive useful insights.
Let’s return to the usage experience, where at least two types of valuable expertise are required: first, the right KPIs must be selected and, in addition, the correct threshold levels must be set for each. To accomplish this, primary customer research is of immense value. Ericsson has conducted extensive customer research where users were able to self-report when their experience was hampered and what they perceived at those times. By matching those quality perceptions with various types of KPIs at particular values, it is possible to determine that when those KPIs cross a certain threshold, customers will likely perceive an impaired experience.
However, that is just half of the challenge. The second task is to correlate those events with detailed network session records to determine exactly what kind of fault, performance issue or event is the root cause of the quality issue. To do this, detailed network expertise is needed to identify relevant types of events. Using a purpose-built, automated, real-time analytics engine to correlate both kinds of data is the key to powerful customer experience insights and associated causes. Detailed network knowledge and actual customer insights are key ingredients in defining and implementing any big data solution that seeks to guide improvement of the customer experience.
A common analytics platform for all
With many different constituencies interested in different facets of the customer experience – from marketing to customer care to the SOC, and so on – it’s a good idea to implement a common analytics platform. That way, each group can leverage the vast and varied data collected in their specific applications of interest.
By adopting a layered architecture, many different objectives can be served while avoiding replication of effort and expense. This layered approach generally includes mediation and data collection at its lowest level. A data correlation layer resides above this, using stream processing, complex event processing and other techniques to assemble session records, user profiles and other useful objects. This can be done in real time, or from historical data. Above that, a variety of business rules engines can be implemented to apply correlations and other rules. These rules engines can share the large volume of correlated source data, and they can be tailored to different types of analysis, such as customer behavior or usage quality.
Finally, various application and visualization layers can be provided to deliver the insights to different internal user groups in the form required. Alternatively, many automated actions based on these insights are possible, such as automated network optimization, automatically generated trouble tickets, tailored service offers and so on.
Big data alone is not enough
In summary, to deliver a quality customer experience across the customer lifecycle, operators must understand how customers view their current experience and what, exactly, they want. Big data analytics, combined with the necessary network expertise, can provide the required insights. Only then can operators implement improvements in their operations that will deliver the superior experience customers desire and also deliver the desired results for the business.