Call detail records (CDR) is a common data source for telecom data analysts to gain insights about mobile subscribers’ behavior or develop features for predictive models.

Example of a typical CDR table:

However, CDRs could be used to extract even more profound insights about individual subscribers.

Here are 3 types of metrics about subscribers that can be derived from CDRs to improve your customer understanding and predictive model accuracy.

Type 1: Advanced metrics for customer activity patterns

To better understand customer activity data analysts typically start by calculating subscriber usage (e.g., Minutes of Use (MOU), MBs of use (MBU), SMS sent) and consumption frequency KPIs (e.g., number of calls made, number of data sessions, etc.). Different variations of these metrics can be developed using information from CDRs (e.g., incoming/outgoing number of calls made, on-net/off-net SMS sent).

Although these metrics provide a better understanding of customer behavior over a specific period, they not always help to quickly identify changes in activity patterns on the individual user level. More advanced metrics like ‘days since last activity’ can help to detect these changes significantly faster.

In simple terms, these metrics calculate how many days a user didn’t make a specific action. For instance, how many days passed since the user made a top-up or browsed the internet. If the gap between the actions is bigger than usual, it could indicate changes in customer behavior, like disengagement or even churn.

Different user consumption patterns visualized using “Days since last metrics”:

This type of metric is especially beneficial when building predictive models focused on predicting customer activity (e.g., will a customer make a top-up in the next 7 days). We also noticed a significant improvement in the accuracy of the prepaid churn models when these metrics were included.

Of course, these metrics need to be re-calculated daily in order to use them in the subscriber behavior analysis or predictive models.

What questions can you answer using these metrics?

  • How many subscribers did a specific action in the last 7 days?
  • How many users did no action in the last 30 days?
  • What is the typical gap of no action for an individual subscriber?
  • Which customers did no action for an irregularly long time?

Type 2: Top cell metrics

The second type of metrics that could be extracted from CDRs focuses on identifying the most frequently used cell in the network on an individual subscriber level.

It is especially useful in analyzing customer mobility and predicting customer satisfaction or churn, as they contain context about a location where a customer is typically using the services. It doesn’t provide exact customer location but an area of the tower where the top cell is located.

What questions can you answer using these metrics?

  • What areas have the largest clusters of active subscribers at a specific time of the day?
  • Which cells most frequently correlate with churning customers (could indicate potential network issues in the area)?
  • In which area subscriber spends most of his time during a specific time of the day (identify home & work locations)?

Type 3: Social network metrics

The third group of metrics can be used to identify subscriber social networks as a group of friends or family. By analyzing the frequency of made calls and sent SMS to specific MSISDNs by an individual subscriber, data analysts can identify which subscribers spend most time communicating with each other.

These metrics could help to decide which subscribers should be engaged with Family deals or how many peers in their network have competitor services (often subscribers migrate to competitor networks used by friends or family members due to cheaper on-net calls).

To sum up, CDR is a crucial source for telco data analyst as it allows to derive deep insights about subscriber behavior and future actions.

However, even for telco data analysts, it is not an easy task to build these metrics, as they need to be updated on daily basis or even in real-time. Also, rigorous data quality and governance processes need to be set-up to ensure high-quality insights and fast reaction to subscriber behavior changes.

Exacaster Customer Data Platform “Customer 360” allows to streamline the metrics development process from any data source and enables the analytics team to develop, test and automate metrics calculation in a few simple clicks.

Learn how Exacaster „Customer 360“ platform can help to improve your analytical capabilities and customer understanding. Contact us for demo!



Senior Data Analyst

Agnius is a senior data analyst at Exacaster using advanced analytics and Hadoop tools to provide meaningful business insights for our customers. More than 3 years Agnius is working with Scandinavia’s leading mobile operator and is a go-to person to gain insights about mobile subscribers’ behavior or features for predictive models.