AWS for Industries

Enable Analytics and Insights for Telecom Networks

Telecom networks generate voluminous fault, configuration, accounting, performance, and security (FCAPS) data. Communication Service Providers (CSPs) are looking at ways to use this data efficiently to improve the network health, as well as reduce Mean Time to Detect (MTTD), and Mean Time to Resolve (MTTR). They want to increase the service uptime, reduce spectrum interference, reduce network capacity forecast, and eliminate drop calls. This business’ desired outcome can be fulfilled by leveraging machine learning/artificial intelligence (ML/AI).

Telecom network operation engineers are not data citizens in general. Although they understand the call flows, KPIs, and metrics, they may need data science skills and require spending time to build highly complex ML models. In this post, we explain how CSPs can leverage the built-in capabilities of Amazon QuickSight to create intelligent insights for their network performance. The dataset we are using for this blog is a CSV file that has the Voice over Long Term Evolution (VoLTE) drop calls count and percentage by zone, location, and tower. The dataset shows also the drop call counts by VoLTE network function such as Evolved Node B (eNB) and Mobility Management Entity (MME). Here’s a screenshot of the dataset:

Figure 1 Drop call datasetFigure 1 – Drop call dataset

Solution overview

In this demo, we’ll share the instructions to build a data pipeline and insights for VoLTE drop calls. We’ll leverage Amazon QuickSight capabilities to build drop calls analysis and dashboards. We’ll enable drop calls forecast and Natural Language Processing (NLP) for a VoLTE network. The following design is for an end-to-end production version of this solution that includes 1\Data migration from on-premises to Amazon S3 using AWS DataSync. 2\Data discovery, preparation, and integration using AWS Glue. 3\BI reports notification using AWS Lambda and Amazon Simple Email Service (SES). 4\Build complex AI model to predict anomalies and root cause analytics using Amazon SageMaker. The portion 3 & 4 are not covered in this blog; we will address them in a future blog.

Figure 1 – Design architectureFigure 2 – Design architecture

Solution implementation

1. Follow the steps outlined to synchronize your data to Amazon S3 to move raw data from on-premise to Amazon S3 using AWS DataSync.

2. Create an Amazon S3 bucket in one of the AWS Regions. When you create a bucket, you must choose a bucket name and Region. You can optionally choose other storage management options for the bucket. Here are the steps on how to create an Amazon S3 bucket.

3. Integrate data from Amazon S3 to Amazon Redshift using AWS Glue. As the solution scales up, there is a need to pull, transform, and join data from multiple sources. To create a single source of truth for all analytics and visualization needs. For Telecom companies, the volume, the velocity and the variety of data will require a fully scalable data warehousing service like Amazon Redshift. We see a need to build an aggregated layer such as one to build KPI for performance drop of VoLTE and also there is a high possibility for large queries to execute in a reasonable amount of time without affecting the Extract, Transform, and Load (ETL) performance. Detailed steps for data preparation and integration with Amazon Redshift are provided in this blog.

4. Connect Amazon QuickSight to Amazon QuickSight. You must create a new security group for the Amazon Redshift instance. This security group contains an inbound rule allowing access from the IP address range for the QuickSight servers in that AWS Region. Here are the steps to connect Redshift with QuickSight.

5. Build the analysis, dashboards, and views on Amazon QuickSight. Amazon QuickSight brings together the efforts of Data Engineering and ETL on the network data to deliver a Telecom Network Analytics Dashboard. Following are the key highlights of the dashboard which includes three sheets (tabs) namely Summary, Details, and Daily Performance.

Summary sheet (Figure 3) includes the network performance metrics. Key KPI metrics like Total VoLTE Calls, Calls dropped and percentage VoLTE dropped are displayed. Users can slice and dice information on the dashboard by Region, Market, Zone or Status either by selecting filters or by clicking on the charts itself. Using Amazon QuickSight’s out of the box forecasting feature, we can easily add forecasts to the trend charts like VoLTE Calls by Day and VoLTE Drops by Day. Amazon QuickSight also provides out of the box Machine learning insights (Key Insights section) that are dynamically generated based on the data. Detailed steps in building QuickSight dashboards can be found in the Author Workshop.

Figure 3 Summary sheetFigure 3: Summary sheet

Details sheet (Figure 4) shows the percentage call drops by Cell Tower and it is sorted descending for the ease of readability. Drop call details data are also displayed for every single tower in the network. Using conditional formatting in QuickSight visual cues can be provided for an easy interpretation of data. % Drop call greater than X % can be displayed as RED. Filters or controls in QuickSight can be used to easily narrow down the data for a specific date range or a tower status.

Figure 4 Details sheetFigure 4: Details sheet

Daily Performance sheet (Figure 5) provides a quick snapshot of daily network performance metrics by zone, market and tower. The map displays cell tower status for all locations. This can provide insights or patterns when a cell tower is down or locked.

Figure 5 Daily Performance sheetFigure 5: Daily Performance sheet

One of the critical challenges with dashboards is that with data changing and new insights being generated all the time, it can rarely answer all the questions. Using Amazon QuickSight Q (Figure 6), our NLP search capability, you can ask business questions in natural language and receive answers with relevant visualizations instantly to gain insights from data by simply typing it in the Q-bar. The following are sample questions that are based on the network performance data that users can ask on the fly:

  • Show me top 10 tower with most call drops
  • Which location has fewest calls in West Zone?
  • Which cell tower in Sacramento has the most % drop calls?
  • Why did drop calls decline for Sacramento on September 29 2022

Figure 6a: QuickSight QFigure 6a: QuickSight Q

Figure 6b: QuickSight QFigure 6b: QuickSight Q


In this post, we went over one of the ways to create a data pipeline to enable analytics and insights for telecom data related to drop calls in this case. We shared the steps to build analysis, dashboards, network drop call reports, forecast, and NLP to make easy for non-business analysts to get an easy and quick information using plain text. The same steps can apply to other telecom related datasets such as Call Data Records (CDRs). To learn more about how telecommunications companies are leveraging AWS Services, visit Telecom on AWS. You Join the Quicksight Community to ask, answer and learn with others and explore additional resources.

Mounir Chennana

Mounir Chennana

Mounir Chennana is a Senior Solutions Architect within the AWS telecom business unites. Mounir has about two decades experience in the telco industry. He’s passionate about telcos network AIOps, OSS/BSS transformation, and leveraging the emerging technologies such as Web 3.0, AI, Generative AI, and advanced analytics to transform Telcos’ businesses and networks.

Mark Ryan

Mark Ryan

Mark Ryan is a Solutions Architect Intern at AWS. He is specialized in data analytics and business intelligence. He’s passionate about telecommunications transformation using the cloud analytics services.

Neeraj Kumar

Neeraj Kumar

Neeraj Kumar is a Senior Solutions Architect for Amazon QuickSight. Neeraj started his career as software engineer building software applications for automotive, manufacturing and telecom companies, he further progressed as specialist and while working at Cognizant he was responsible for designing and developing end-to-end Business Intelligence and Analytics solutions for major Insurance companies.

Salim Khan

Salim Khan

Salim Khan is a Specialist Solutions Architect for Amazon QuickSight. Salim has over 16 years of experience implementing enterprise business intelligence (BI) solutions. Prior to AWS, Salim worked as a BI consultant catering to industry verticals like Automotive, Healthcare, Entertainment, Consumer, Publishing and Financial Services. He has delivered business intelligence, data warehousing, data integration and master data management solutions across enterprises.