Menu
OVERVIEW
What was the project about?
FMS (Fraud Management System) for the Asset & Inventory Management business of AT&T Wireless handles the following scenarios enabling predicting patterns and analytics on AT&T Inventory.
- Devices & shipment tracking with their state patterns determined
- Analysis and finalizing risk zones and fraudulent entities based on regression data
- Aggregation on Robocall data identifying trends & patterns for fraud prevention & aid solutions.
TechniKons designed, developed, and implemented FMS application, It is a 2 years SOW project. TechniKons’s project services team met the immediate and ongoing needs of AT&T to develop the FMS application to focus on its greater business objective.
Company Name
AT&T Wireless
OVERVIEW
AI / ML Model Development
Location
Plano, TX
CHALLENGE
What was the main issue(s) you and you team were trying to solve?
FMS (Fraud Management System) is a Risk Management Platform offering fraud detection & prevention services to AT&T’s Businesses. AT&T had a few peculiar problems listed below,
- Robocalls & related frauds had increased multi-folds, leading to high complaint volumes and negative customer experience.
- Apple iPhones & iPad shipment theft were on the rise with monetary & reputation losses mounting on every shipment theft.
- Device loss had been a growing issue from the regions across post-sales and returns leading to expensive reactions from AT&T’s end.
For the above problem statements, TechniKons was requested to provide a prediction solution to deter any further losses.
SOLUTION
What was the solution that was implemented for this?
TechniKons, experienced in constructing ML Models, created a framework for data ingestion, exploratory analysis & Model performance operations.
TechniKons built a Regression-based Prediction model to determine and proactively deter any Fraud-enabling patterns and analytics. It was a cloud based solution worked along-side with multiple products in Microsoft Azure like DataBricks & Relational AI to aggregate data & design Prediction models.
Post Model creation Technikons worked with AT&T Data team in training them with Data Quality Management and Model Operations & maintenance to improve the model with every new node addition.
RESULTS
How did the Solution fix or improve the company’s problem areas?
With Fraud prediction solutions deployed to FMS, it aided AT&T with the following outcomes.
- Upon a period of 1 year the Robocall assessment and blocks were improved with Robocall complaints dropping 23%.
- For Apple devices shipment thefts, our IMEI-based geo-position analysis highlighted the patterns of shipment movement to map out fraud intent patterns and respond to them proactively, deterring theft.
- Using Device sales/ returns & Lost report patterns, TechniKons were able to provide Hot zones & flagged parties marking High-risk behaviours causing losses. Upon implementing targeted measures, the hot zone losses were reduced by 30%, thus saving money & reputation.
