Why is there a delta between the measured network performance and network quality perceived by the customer?
Rolling out and scaling a resilient 5G network
COVID has brought networks across the globe into sharp focus as employees across the globe have had to resort to working from home. In addition to this, more stress has been put on the network with the growth in remote education, mobile gaming, increased streaming of HD and 4K video and the emergence of 5G. In the ultra-competitive Communication Service Provider (CSP) environment wherein subscribers can easily switch their service providers CSPs’ need to provide a quality of experience (QoE) that minimises possible churn and provides a competitive advantage to the CSP. For the CSPs’ this has meant an increased focus on ensuring the current networks can cater to the increased bandwidth requirements whilst ensuring good customer experience through AI-powered intelligent network operations. This, along with the ongoing rollout of 5G, has resulted in a renewed focus on network analytics to ensure :
- Improvement in customers’ network quality of experience
- Effective planning and management of dense networks
- End-to-end automation across the network
The ability to use network resources effectively and predict faults/alarms before they appear gives a pretty good measure of the maturity of the CSP w.r.t. Network Analytics. The primary use cases CSPs are focusing on to improve the maturity of network analytics include:-
- Speedy detection and prediction and resolution of network errors: Proactive and early detection speeds up the process of identifying the source of and resolving network faults. This use case also includes predicting degradation in Quality of Service (QoS) to reduce potential customer-impacting problems on the network backbone. This is achieved by utilising predictive maintenance and anomaly detection. E.g. for predictive network maintenance using anomaly detection, the data utilised include real-time information about network element, network outages and network performance degradations alarm/alert data along with historical network element Data and ticket history.
- Effective forecast of network traffic: Load Balancing is essential to optimise new infrastructure planning, optimise network asset utilisation as well as transitioning from legacy to next-generation network solutions to ensure quality and consistency. E.g. the forecast for traffic and utilisation can be derived by utilising sales and marketing forecasting data, along with site-level data like demographic, content and usage data whilst considering seasonality and event impacts.
- Network optimisation: There is an increased focus on CSPs’to Optimise the use of network resources to mitigate the impact of network faults and adapt to or anticipate changes in demand. Thus network optimisation is fundamental to automated service provisioning in software-defined networks, and providing insights on network usage and traffic, which are essential for new service creation.
- Optimal customer experience: The challenge is that while there are models to predict overflow of traffic and balance traffic at the device level, CSPs have not been able to do the same at the service level in a smart, cost-effective way without undermining the customer experience. The network needs to scale to meet customer requirements ensuring improved NPS. The subscriber usage and mobility patterns together with customer information and complaints data can be utilised to understand better the gap between the measured network performance and the quality of the network perceived by the customer.
Besides the above use cases for using Artificial Intelligence in a network context, there are several other use cases spread out over access networks, core and NOC & SOC. The machine learning technique to be applied for the analytics can be chosen based on the use case and the availability of historical data. Supervised Machine Learning models are most suitable for use cases where historical data is available; if the data is not available, then the unsupervised machine learning models can be utilised. The machine learning models can be further augmented by using natural language processing techniques for unstructured data generated by end customers via Social Network activity and freehand text from the interaction through various channels. The use cases mentioned above and other use case are being brought together by CSPs under the theme of autonomous networks to achieve rapid scale-up, simplified operations, development and maintenance and self-healing of the network.
Advanced Analytics, AI and ML platforms enable CSPs’ to add value to their organisation by using advanced analytical techniques and AI on network and customer data together. These platforms provide the ability to integrate customer experience, equipment, external and financial data with network traffic data and analyse the mix to provide the ability to make faster and more automated decisions to drive down CAPEX & OPEX and at the same time find new revenue opportunities and increase profit potential.