Using Telecom Data, Device Data to build credit scoring ( To Lend For Personal Loan, Installments…)

A variety of indicators from the telecom industry are used as an input feature to predict customer attrition. Big Data Scoring is a cloud-based telecom data decision engine that helps banks and consumer lenders improve credit quality and acceptance rates through the use of big data.

Sampling techniques have been evaluated based on Logistic classifier, Random Forests, Neural Networks, GBM, … Boost conversion with ML and pre-approved leads. This help to optimized decision system.

Churn is defined as the propensity of a customer to stop doing business with an organization and subsequently moving to some other company in a given period. In telecom industries, Customer attrition is becoming a significant area of concern.

Maintaining overall profitability over a long time demands current customers to be always with an organization; hence, it is indispensable to retain the customers; therefore, it becomes essential to know churning customers beforehand. Losing customers not only lead to a loss in revenue but attracting new customers also involves a considerable amount of investment further impacting the income of the organization.

Developing and deploying custom scoring models that combine a lender's internal data with thousands of pieces of external data. Machine Learning to model variables to pick micro patterns which include high dimension Feature Space such as: Demographics, Data Usage, Social Network, SMS, Top-Ups, VAS, Calling Behavior, Geo-Location, Data Usage, Apps Usage.

Telco data

How does Telco data scoring for individual personal finance ( as cash loans, installment loans, …) work ?

Consumer: For individuals’ who do not have bank account or finance records, Telco data is useful in predicting their risk level. That from make use of Telco data and machine learning to boost predictive power and increase acceptance for small and micro loans.

Entry point to credit inclusion

Attributes collected from Operators for Risk Models

Illustrative list of variables:

1.Pre Paid/Post Paid

• Information of Activation of Sim card or Tenure of customer along side other details mentioned in the application form
• Postpaid defaults, credit, churn and payment information. Availability of mobile wallets

2.Data Usage

• Data used, revenue generated from data, hourly usage of data, data related value added services
• Additional details on applications used, websites browsed over day and night

3.Geo-Location

• Mobility from CDR data, day and night presence, density of location
• Prominent location with attributes from census, public available data

4.Top Up history

• Top-up information, type, size and frequency of top-up along side channel of top-up
• Channel of bill payment, invoiced amount, payment terms, and mode of payment — Bank, Credit, Wallet information

5.Calling/SMS Patterns

• Statistics on Call duration/count, calling phone numbers, towers, SMS sent and received
• Time Of Day calling, weekday/weekend, inactivity, calling consistency

6.Demographics

• Demographics information recorded at the time of customer filing for sim-card
• Inferred demographic details

Some out of the box attribute creation

Calling behavior patterns correlates to risk . We transform the raw CDR (Call detail records) into behavioral patterns to correlate with risk.

Network Analysis

Depending on the network of the individual we track the change in trends over months

Customer Journey Lead Generation through Telco Scores
Lead Generation through Telco Scores ( As the new customer for bank to explore)

Device Data

Consumer to need individuals’ who do not have bank account or finance records. Mobile data is useful in predicting their risk level.

Make use of Mobile data and machine learning to boost predictive power and increase acceptance for small and micro loans.

Potential to credit profile more than 5BN consumers globally

Harnessing the power of ‘ready-to-use’ Alternative Data

Device Data can assess the creditworthiness of applicants who are unable to obtain credit via the use of traditional data:

• Taps into one of the richest sources of alternative data to predict credit risk
• Quick ROI as solution is scalable, available globally and can be implemented within days
• Seamless customer experience through automation. No manual intervention
• Stringent processes to ensure maximum customer privacy and data security

Predictability of Device Data. Use of ML over rich attributes from device prove to be extremely powerful to predict risk.

Architecture for Scoring Device Data

Server less architecture on cloud allows to generate scores in realtime and handle scale

X Score Alternate Credit Score for lending

Problem Statement

Solution

X-Score enables credible information access to banks for credit decisioning

Score Pooling

Alternate Data Projects

Proven capability in risk scoring various verticals in emerging markets

GINI — Measures the rank ordering or performance of a model to predict risk

Decisioning lifecycle for a digitally enabled bank

Decisions across the customer lifecycle

A better customer journey

Business impact can be tremendous