After reading Eric’s post yesterday, “Why Revenue Assurance Should Embrace Data Science”, I thought about my own work and how we do revenue assurance and fraud management at Epic.
My daily work consists of writing SQL queries and checking the results of controls that are setup as SQL procedures, Tableau data visualizations or even jobs in (the community version) of Pentaho. Our BI and IT team pulls the raw data from our systems — charging, billing, as well as all network elements. Upon this layer I build controls as SQL procedures which utilize the SMS and email sending capabilities of our data warehouse to alert of possible fraud or exceptions, and to feed data visualizations in Tableau such as the below:
What you can see here is a very simple control that compares the outbound roaming duration from charging and TAPIN. This is a snippet; there are filters to the right of the actual data visualization which lets me filter by VPMN or date.
There are lots of controls running that send me alerts; for fraud these include NRTDRE checks, wangiri and PBX hacking. There is an ongoing and continuous effort to ensure some high risk items (e.g. premium numbers, SMS short codes, roaming) are not abused. Trends of all network elements and the charging system are in Tableau with filters to dynamically change the views. In some cases, the two systems can be compared on an aggregate level such as in the above example.
In each telco setup you can identify high value controls that will assist your department. High-value controls might be assuring the reports that are used as inputs to your ERP. Assuring the revenues inputted should be bread and butter to a Revenue Assurance department.
Margin analysis is another high value project that can be completed when all data sources are available and combined in the database. This can also guide the thresholds, or even creation, of some controls and alerts, as well as assisting commercial departments.
The data analysis skills to use SQL programming has been complemented by the telecom knowledge I have gained from various sources, including online courses, talking to people, reading Commsrisk, and using resources from RAG ,. This combination has enabled me to become a data analyst that can take data and not only provide a data extract/report but also inject some insight on whether we are losing money.
Given my experience, I can relate to the data science/analytics part of what Eric mentioned in his post because that is the method I use. I am not a data scientist per se, since I do not use stats or machine learning, but I build controls using SQL and Tableau to do revenue assurance and fraud management. It is my belief that, given enough resources, statistics and machine learning can also be built and maintained by RAFM departments.
For me, revenue assurance is one and the same as data science and analysis. In fact, one of the prerequisites for data science is domain knowledge. Revenue assurance and fraud analysts can be very valuable within a telco if they can define themselves as data analysts or data scientists.