Telecom operators of today generate vast amounts of data coming from the use of services such as voice calls, SMS and mobile Internet. This data can be exploited using Machine Learning (ML) for intelligent analytics to tackle problems such as identifying network anomalies, forecasting network performance, identifying potential customer churn, predicting the next best offers and optimizing the Customer Relationship Management (CRM). In this post, we’ll talk about a few of these use cases and how Telcos can use Teradata Vantage
to tackle these problems.
Automatic network anomaly detection.
Ensuring good Quality of Experience (QoE) is essential for Telecom operators to prevent any churn. Network performance is usually monitored manually by engineering teams consuming a significant amount of human resources. Automating this task can significantly reduce the OPEX for any mobile operator. This automation can be accomplished using ML, where historical network data can be used to identify any current or future network anomalies and in turn, can be used to isolate the cause of the anomalies as well.
Intelligent network dimensioning.
Network operators are constantly faced with the challenge of ever-increasing mobile Internet traffic. Catering for this enormous growth of mobile data requires efficient network planning to forecast any future network expansions. ML/time series analysis can allow for accurate forecasting of network traffic into the future thus allowing operators to plan any capacity expansions to ensure the best possible QoE for their end-users.
Predicting customer churn.
In a highly competitive market, operators need to keep their customers satisfied to ensure that the business remains profitable. Predicting customer satisfaction is made possible using ML where the subscribers' service usage pattern allows them to identify a possible churn. Early detection of a potential churn can allow operators to take pre-emptive steps to prevent the subscriber from abandoning the service and even optimizing the network through root cause analysis.
Modern Telcos usually propose a wide range of service offerings. Predicting the Next Best Offer (NBO) is a standard technique employed by content providers to identify the most suitable content for a potential subscriber. Using the service usage data, operators can build subscriber persona profiles that can allow them to intelligently know the best offer for their subscriber base. This means that the most relevant offer is advertised to the customer, which in turn maximizes the chance of the product uptake.
Intelligent Customer Relationship Management (CRM).
Customer management is an essential element of any business. Dedicated CRM teams often work without any knowledge about the background of the target customer. ML can be used for subscriber profiling which can quantify “customer value” that allows efficient customer management for current as well as future potential customers to drive sales growth.
The use of ML for each of the above-mentioned use cases requires access to clean and detailed network-level and subscriber-level data. The network-level data often comes in the form of Key Performance Indicators (KPI) generated by the network entities. While the subscriber level data comes from the Call Data Records (CDRs) that log the subscribers’ service usage. Most often, different business units in a Telco work in silos with different sets of tools for analyzing the datasets and/or building the predictive models. For example, a dedicated DWH team will be responsible for only storing the data, while another team will be responsible for analyzing the data. Passive data warehouses and the diversity of the tools used by the different business units result in data silos within the company that hinder the organizational performance. To tackle this problem, data warehouses need to get intelligent enough for data analysis and eliminate the need to remain static data dumping grounds.
achieves this goal by combining data warehousing, analytics and machine learning into a single offering, which means that any business unit can easily and quickly make the best use of the available Telco data
. A unique feature of Vantage is its ML engine that allows the building of ML models quickly in the database
without the need for any external data extraction or preprocessing. This significantly speeds up model testing and reduces the time for the models to be deployed in production. With Teradata Vantage, any resource can perform intelligent, predictive and prescriptive analytics on Telco data for their specific tasks. For example, a network engineer can get network traffic forecasts for efficient capacity planning, or a data scientist can quickly build churn prediction models for ensuring customer retention. In summary, Vantage allows Pervasive Data Intelligence
, where any business unit – be it engineering, marketing or customer support -- can benefit from ML for optimizing their respective workflows.