Machine Learning to monitor, optimize & assure Quality of Experience
Customer experience has been and is a battlefield for telecom market share. Mobile video consumption rises 100% every year (Insivia) and will account for nearly 80% of mobile data traffic by 2022. The cloud gaming market is reaching $8 billion in value by 2025 and the largest driver of growth across the video game industry is from mobile gamers. End users have less and less tolerance for QoE degradation – jitter/buffering/freezes etc.
If mobile operators are to convert the promise of 5G into new revenues from new services, they will need to deliver the consistently high network experiences which consumers have come to expect. But providing quality experiences – at a time when network technology is going through unprecedented levels of transformation with operators inevitably bearing a greater responsibility for the successful operation of the network – will not be straightforward.
Approaches to network quality and network experience need to change because of the changes in market focus in 5G.To help the CSP address the challenge, Nokia developed its AI for video-based services solution. This analyzes video/gaming performance data including encrypted video, to predict how Quality of Experience (QoE) will be affected by interference, congestion and coverage. Modeling techniques then link video/gaming QoE to critical business Key Performance Indicators (KPIs), including churn, Net Promoter Score (NPS) and revenue. The CSP is provided with recommendations on which parameters to alter for resolving issues quickly. Linking QoE to business performance also allows optimization efforts to be directed at the most valuable subscribers and locations. Correlation of User Plane traffic and Layer 3 signaling enables prediction of ‘freezes’ and QoE. The solution is multivendor capable and can make use of a variety of data sources. Nokia offers flexible commercial models. For a US CSP, the solution made 54 automated recommendations, allowing it to optimize the performance of the network by changing parameters. The worst performing cells were highlighted, as well as issues affecting the transport, core and content delivery networks.