Network consumption patterns have dramatically shifted over the last few months as a response to social distancing measures, changing in volume, geography and time of day. Many companies have implemented work from home policies. Schools are closed and transitioning to digital learning. In addition, sporting events, movies and other entertainment are not available, resulting in increases in streaming services, gaming and other “remote” entertainment. Communications companies want to proactively manage resources, including their networks and workforce, to maintain customer satisfaction. A variety of analytics related to network monitoring, network planning and proactive maintenance are opportunities for network providers to leverage advanced analytics to support all needs for the network.
Network providers can use analytics to predict the network use and recommend appropriate augmentation based on the needed bandwidth. You can base analytic models on very granular network data at the tower or node level to enable pinpointing geographies that require specific augmentation or optimization. Providers can also use analytics to monitor the network. Identifying anomalies and capacity issues can be an opportunity for proactive intervention or prioritization. Resources are scarce, and using analytics helps optimize investments in augmentation (whether they are equipment upgrade plans or mobile cell towers).
With network augmentation analytics, teams build predictive models at very granular levels, such as node or tower. Then they link them together using the network hierarchy. The result is a large number of models linked by a hierarchy. Network analytics teams need to leverage advanced techniques and model automation with these linked models to create models that are highly accurate in predicting future bandwidth needs. By predicting network requirements by location, they can identify specific locations where demand will exceed the threshold for capacity.
Using a decision engine built around the providers’ specific business protocols that incorporates these models, network teams can quickly assess model results, recommend needed augmentation and prioritize locations. Then they can easily identify areas that require improvement. Teams can easily review results with a visualization tool, enabling quick decision making and prioritization. By leveraging model automation and decisioning, teams can accelerate the time to results.
Monitoring the network can leverage advanced analytics as well, enabling communications providers to be proactive in addressing potential issues with their network. Analytics makes it possible to monitor network devices using a variety of techniques. Anomaly detection techniques in use today range in complexity. One popular technique is support vector data description (SVDD). Additional techniques include ARIMA models and regression.
The benefit of an analytics environment for network sensors is that models can be built using different techniques and can all be put into production with real-time sensor monitoring. Leveraging real-time tools capable of alerting, network groups can run a variety of models to monitor network sensors and be alerted when conditions are out of the normal expectations. Analysts can review visualizations of the real-time data upon an alert to diagnose. With the detection of a potential network, analysts can focus on higher risk areas with quicker responses, resulting in higher customer satisfaction.
In addition to using network data to predict how consumption patterns will change, providers can mine customer care call and ticket volumes, as well as ticket information or content, to find ways to improve customer experience. Customers contact providers to report issues, and providers can use this information to proactively identify outages or other large-scale issues.
Using call and ticket volume patterns
We can apply the anomaly detection techniques discussed above to monitor ticket volumes incorporating type, time and day to identify when the number of tickets is getting outside of the normal business range. The system can send alerts to business analysts who can use visualization to identify potential large-scale product or service outages. Getting ahead of larger-scale outages can result in higher customer satisfaction, as well as support fulfillment of SLAs.
Providers can also optimize staffing with detailed information from ticket and call volume. Detailed forecasting, incorporating events and other information, will produce highly accurate forecasts by time of day and type of calls. These forecasts can optimize the allocation of staffing levels and skill sets to support care center volume and type of call within the standards of customer satisfaction without overstaffing.
Using unstructured data
Providers can gain insights using text analytics to mine unstructured ticket notes within trouble tickets to uncover root causes, as well as the need for new processes that should be created. For example, comparing topics derived using text analytics and ticket categorization can be enlightening. Situations where the text reflects a specific issue (topics categorized using text analytics) and the ticket is categorized differently represent opportunities for new protocols or additional education, all with the intent of speeding time to resolution or mean time to repair.
Text analytics can also provide benefits by informing chat and chatbot intents. Providers can mine transcripts of chat encounters with text analytics to identify themes. They can use the resulting insights to enhance agent-assisted chat. You could also perform text mining on this unstructured text to identify the most common chat topics to inform chatbot intents. By continually enhancing chat experience with text mining, providers will maintain customer satisfaction.
Intelligent service assurance
Communications companies also benefit from process automation to resolve issues before customers are even aware of them. Proactive resolution will reduce the number of care tickets and subsequent workload.
Providers can design and deploy workflows to automatically resolve emerging network issues in real time. For example, with a fixed-line network, you can monitor and evaluate routers for automated actions to resolve network concerns. Providers can enrich and aggregate streaming sensor data to report trending key performance indicators. You can also build models to predict the probability of a variety of needed actions. When the system predicts a high probability of a particular problem, the provider can automatically push actions such as router restart, or send a software update to the router.
Providers can track and monitor automated actions to take additional actions based on history. For example, if devices consistently require restart, or restart does not resolve the issue, more deliberate action can be taken, such as sending a new router or a technician for repair.
Providers can accelerate the issue resolution process with a full analytics life cycle from data, modeling and deploying recommendations. They can take action before the customer is even aware of the problem. This maintains customer satisfaction and reduces reactive troubleshooting, freeing up key resources to focus on complex issues.
The long view
With unexpected shifts in network consumption, providers want to be agile with monitoring to ensure they are able to detect network and equipment issues. Assessing capacity planning, network operations and service assurance can be enhanced with analytics. These advanced analytic techniques and automation can help providers continue to maintain their networks and customer satisfaction.