Next Generation Customer Experience Management systems

Big data and analytics are a great challenge for telecommunication operator to automize and simplify network control and operations and improve customer experience

The combination of increasing network complexity (4G, 3G, 2G, WiFi, Femtocell, NFV/SDN …), with the exponential growth of data traffic and the need to support OTT services carried by their networks, are overwhelming the Telecom company’s ability to provide high-quality service assurance to their subscribers.

• Telecommunication providers don’t control anymore the end-to-end services as most part of these services are provided by external companies (f.e. OTT). But at the same time, subscribers still credit Service Providers for the end-to-end service.
• The coexistence of different network technology – 4G, 3G or 2G, WiFi … has dramatically increased the complexity of measuring service quality.
• The User interaction is completely changed too: subscribers and their devices (more than one) are always connected to the network (always on). Customers are become extremely demanding in term of quality and speed of the connectivity.

Telecom operators need new generation of Service Assurance systems – based on advanced analytic capabilities – to automatize and simplify network’s control and operations for preventing network’s issues from impacting subscriber’s Quality of Experience (QoE).

In the old telecommunication scenario, the quality of the services was tightly bound with the network quality: when a network element fail, a network alarms are raised. Through the mapping between service and network elements it is possible to identify services impacted. Similarly, when a service KPI/KQI is degraded (it goes above or below defined thresholds), a KPI violation alarm is raised and through the service-network mapping is possible to identify the cause of the problem. So the combination of alarm, KPI and service mapping allows to manage problem and incident in the network and/or in the services provided to the subscribers

In the “4G and Apps era” the equation: Good network SLA = Good Services is not anymore valid. The traditional network-service scenario is completely reversed: only few services are still network services; the most part of the services are provided by application over the network (the so called Over-The-Top applications) and in some cases are also device-dependent. Telecom operators don’t control anymore the end-to-end service and they need new tools to manage subscriber’s QOE.

The new assurance systems must be a combination of descriptive analytic – to analyze the status of the service and the subscriber’s QoE – and predictive and associative analytic – to automatically identify the relationship between QoE violation and network behavior and recommend corrective actions.

The Customer Experience Assurance (CEA) system will be to able to: 

• Measure and monitoring service QoE, collecting data from all Telecom operator infrastructures (control and user planes, devices, server …) and aggregating them to produce QoE KPI/KQI.
• Automatically and dynamically associate QoE KPI/KQI information with networks and infrastructure data to identify the relationship (and the root cause) between QoE violation and Operator infrastructure alarms
• Recommendation analytic methods, in order to automatically provide the corrective actions to solve the identified issues.
• Predictive analytic methods, in order to predict future issues and to support recommendation analysis.

as represented in the picture below:

The associative engine and the recommendation (or decision trees) engines are the core of these systems:

• Using associative statistical techniques is it possible to identify hidden relationships between different type of data (f.e. devices status and network alarms) and and the relevance of these relationships from QOE management. They allow to automatically identify:
  • IF a specific QOE violation is related to specific network condition (f.e. download speed drop on terminal type A when radio signal power is medium in Los Angeles stadium area).
  • HOW this specific relationship is relevant in terms of frequency (f.e. how often these information co-occuring simultaneously)

In other words, for every QOE violation event the system looks to the whole CSP infrastructure alarms to find the more probable associations (“rules”) between these elements. All these associations are grouped by similarity to identify the more relevant an the more important.

The recommendation and machine learning models are used to suggest the best corrective action (often called: Next Best Action – NBA) to perform on network based on the root cause analysis performed by the associative analysis engines. Decision trees are the most common techniques used to implement NBA. They are created using known cases and their recommendation efficiency is improved by feedback loops that computes the effectiveness of the recommendations and retrain the model.