An AI use case for reducing revenue leakages

Today, digital transformation is a top priority for service providers. Many of them also realize the potential of Artificial Intelligence (AI), particularly in areas like service and revenue assurance. But how can AI help prevent revenue leakages in an increasingly complex application landscape? Find out, here.

An AI use case for reducing revenue leakages

Service providers’ OSS and BSS application stacks are becoming increasingly complex. Between delivering the service to collecting the cash, there are literally thousands of problems that can arise. On one side, an error – for example, overcharging – may result in a bad customer experience and a higher customer churn rate. On the other hand, undercharging may impact service provider revenues. TM Forum’s Revenue Assurance Survey Report 2019-2020 puts the estimated global telecom industry revenue leakage at 1.5 percent of overall revenue.

The main issue with revenue assurance is that operations teams need to investigate any issues reactively. Finding the root cause of the problem in bespoke multi-vendor OSS and BSS applications is a complicated and time-consuming process. Revenue leakage could be caused by a simple issue, like a slow Unix server on which billing application is hosted, or a highly complex problem like an incorrect tariff plan.

Reducing revenue leakages with AI: a case study

With help from Ericsson’s AI-powered Operations Engine solution, a Tier-1 service provider in the Middle East with 12 Million active subscribers recently reduced their revenue leakages. Ericsson manages 80 multi-vendor legacy bespoke applications and 390 databases for this service provider, which are hosted on 2,000 Unix servers in four data centers.

Revenue leakage avoidance is a big challenge, and like many service providers, this customer’s revenue assurance team wanted more insights into revenue leakage and if possible, plug those holes.

Call detail records (CDR) get generated every time subscribers use a service through their phone. These CDR files are then sent from the network to the mediation system, which sends these files to the billing system responsible for processing CDRs and charging the subscribers for the usage based on their agreed plan.

Revenue assurance experts observed a leakage of CDRs between multi vendor mediation and billing systems. Each successfully processed CDR translates into revenue for the service provider, but the billing system was dropping and suspending some CDRs – which created a direct revenue leakage.


An AI-powered solution

The operations team needed a near-real-time solution to proactively identify the CDR leakage. The solution was also required to help find the real problem behind CDR droppage and suspension. This is when the Ericsson Operations Engine team decided to harness the power of data and machine learning to solve this challenge.

Data engineers collected log data from mediation and billing applications as well as the underlying infrastructure on which these applications were hosted. Data was also collected from mediation and billing APIs, and data scientists built machine learning models to find anomalies or outliers in the data. For example, whenever response time of any billing API is outside the normal range, the machine learning model flagged an anomaly. Similarly, an issue in the infrastructure layer like abnormal CPU utilization on a Unix virtual machine was also flagged as anomalous behavior. Based on these anomalies, the operations team took actions like reconfiguring the billing API or fixing the Unix virtual machine.

The same team is now working towards implementing closed-loop automation using machine reasoning to solve the underlying problems without human intervention. All of this resulted in a reduction of CDR leakages between mediation and billing systems. We achieved close to 99.999 percent accuracy in billing based on the reconciliation of CDRs between mediation and billing systems.

An AI-powered solution

 

Future AI use cases for IT systems

As part of the continued focus on AI in operations, Ericsson launched its AI-based Ericsson Operations Engine in January 2019. This particular use case is only one of the 45 AI cases built by over 1,000 Ericsson experts specializing in telecoms and data science. Ericsson managed services now has over 1 billion managed subscribers, and we manage applications catering to over 100 million mobile recharge requests on daily basis, collect usage and generate bills worth millions of euros per month.

Future AI use cases for IT systems

 

To learn more about how Ericsson managed services leverages AI for Cloud and IT service operations, please follow this link.

The Ericsson Blog

Like what you’re reading? Please sign up for email updates on your favorite topics.

Subscribe now

At the Ericsson Blog, we provide insight to make complex ideas on technology, innovation and business simple.