A recent survey by McKinsey & Company of 46 telco CTOs around the globe indicated that the second highest priority, after leadership in the local market, for the rollout of 5G is improved customer experience. This indicates that, within operators, the teams responsible for long-term technology strategy are clearly focused on customer experience as a key differentiator. For a long time, improving customer experience was primarily the focus of the marketing and customer care teams, but it is now the responsibility of everyone in the CSP organization.
In addition, it’s clear that the traditional network operations center approach of monitoring the network through element management systems, performance counters, alarm management and trouble tickets is no longer aligned with the goals of most operators. The traditional approach is based on a break-fix model and therefore is completely reactive. This approach also provides no visibility into the impact of network outages to the services being used on the network and, more importantly, the impact to the customers using those services.
A new approach is required that doesn’t focus on the performance of the network from the perspective of the network, but instead focuses on the performance of the network from the perspective of the customers using the network and corresponding services. In the highly competitive environment in which operators operate, it’s key that the customer is put at the center of every tactical and strategic action and decision.
Steps to creating end-to-end QoE visibility using AI and analytics
Customer experience management systems have typically used the approach of ingesting probe data or network key performance indicators (KPIs) and using this information to generate more network KPIs on a per customer, device, location, or similar basis. However, this approach is purely an extension of traditional network monitoring approaches, where you measure the customer experience from the perspective of the network.
Throughput or latency KPIs tell you how good or bad the network is performing when delivering content to a customer, but they give you no indication of what the actual experience of the customer was when accessing and using a particular type of content or service. These KPIs are only one in a whole range of KPIs that, when combined together, using analytics modelling techniques can give you the real perspective, or QoE, of what your customers are experiencing.
Fundamental to measuring the QoE of customers is leveraging the variety of data available from the network nodes, monitoring and support systems. The volume and variety of data generated by and available to an operator is staggering -- from RAN call traces and measurement reports, core and IMS network transactions, OSS counters, alarms and trouble tickets to billing, provisioning and CRM, drive testing and crowd sourcing, and third-party data like weather and traffic patterns.
Most operators have made some attempts to measure customer experience. The challenge is that often the approach has been fragmented, focusing on either particular services, such as VoLTE or video, or a specific portion of the network, such as the RAN or core. The approach has not provided end-to-end visibility of QoE.
In order to provide end-to-end visibility of QoE, the first task is the complex process of stitching together all of the available data sources. This is a classic big data analytics problem and can rely on:
- Where strong linkage exists, leveraging unique ids that are available between the data sets to be stitched.
- Where weak or no linkage exists, analytics techniques can be used to cluster data sets together
The resulting master flow record must then be classified based on the type of service being accessed, that is voice (including VoLTE, VoWiFi, etc.), video, data and IoT. Once the service is understood, the master flow record can be processed through the relevant analytics model in order to score the QoE for each individual customer per service. The QoE modelling combines the dimensions and attributes available and takes into consideration historical QoE to build a per customer experience profile.
Once the customer experience score is determined, each customer can be automatically monitored to determine any unusual or anomalous variations. In addition, per customer QoE scores can be aggregated -- for example per device or device attribute, per location and per network element -- to provide the customer’s QoE profile, as an individual or in a customer grouping, for each of these dimensions.
The key when anomalous QoE behavior is detected is then to automatically troubleshoot and pinpoint the cause of the anomaly, relating it to a particular group of customers or devices, a service degradation, or a network problem. This can be done by drilling back down into the datasets that were used to generate the QoE score in the first place in order to look for patterns or relationships that indicated where the problem originated. This end-to-end approach enables an operator to understand the relationships between the QoE anomaly, the network elements and devices delivering the service, and any service degradations.
This information can be used across the organization to:
- Predict and prevent service degradations and network incidents before they occur.
- Reduce the noise of alarms and tickets that the network operations team needs to monitor and investigate so they only focus on the customer-impacting problems.
- Enable proactive customer care -- to reach out to customers or fix their problem before they call customer care.
- Personalize customer offers and services based on a customer’s per service QoE (The blog “Four ways CSPs can use AI to gain new subscriber insights, out-market OTT competitors and deliver a better customer experience” by Nicolas Carré outlines this further and provides some excellent examples).
More importantly, operators can model the relationships between QoE for the different services they offer with customer sentiment in the form of churn, net promoter score, or calls into customer care. An operator can understand if poor QoE is the root cause for customers to churn to their competitors or to be detractors. Once this relationship is understood, an operator can react to improve the factors impacting customer experience that need to be improved. In addition, the operator can use this information to predict which customers will call customer care or which customers are the silent detractors who will never call customer care but have a QoE profile of those that do.
Ultimately, this end-to-end QoE approach can be used to drive automation and closed loop actions back into the network, to fix problems before customers even know they have a problem, to automatically raise incident trouble tickets providing clear visibility of the affected customers and root cause of the problem and reach out proactively to customers who operators know are having problems.
To realize value quickly from any QoE approach, operators require out-of-the-box use cases for customer care, operations and marketing that can be implemented and integrated into existing business processes. However, it’s also important that they are dynamic to remain one step ahead of the competition and local market and industry trends. Operators will need to be able to accelerate the realization of new use cases rapidly and be enabled to build their own use cases where appropriate. This is because the traditional approach of monolithic product stacks, with fixed integration points and features, is in the rear-view mirror. QoE use cases will need to be spun up quickly, may have a short lifetime with a defined business objective, and will be augmented or replaced with new use cases as required.
Putting the customer at the center of all business and operations decisions will be the key to future success for operators. This requires AI and big dataanalytics to make sense of the noise being generated by the network in order to zoom into the problems that really matter and impact customers. The flexibility to be able to quickly test new QoE related use cases, validate the results and business case, and quickly roll these new use cases into production will also be a key differentiator for operators.
Finbarr Travers is chief technologist at Guavus (a Thales company)
Image: iStock Photo