Financial services

How to improve the operation of a payment provider?

A strategic study at a large payment provider formulated the following objective to improve returns

To boost revenue from a specific market segment, focus on improving collections of outstanding debt, repricing less sensitive segments, and retaining high-value customers at risk of churn. However, data from these segments is highly fragmented and of vast scale.

The Challenge

Solution

Increase revenue from a specific segment of the market by better collections of outstanding debt, repricing fewer sensitive segments, and retaining high-value customers at risk of churn. The data from the segments were very fragmented and of enormous scale.

Existing Infrastructure could not run analyses necessary to achieve objectives.
No effective way to prioritize or rank merchants for collections
Limited ability to determine the best rates and fees for a merchan

Complete view of merchant activity — The company developed an unprecedented understanding of their customer base with a large volume of integrated data on customer activity, pricing, payment terminals, billing, fraud, and credit history.
Improved collections — Analysts develop statistical models to rate accounts by how likely they are to pay. High-likelihood accounts are routed to internal collections teams, while low-likelihood accounts are referred to third-party collectors.
New pricing strategies — Analysts perform high-scale analysis to assess the impact of different fee structures on customer retention. Sales teams use these insights to better price new accounts and reprice existing accounts to prevent churn.

Existing Infrastructure could not run analyses necessary to achieve objectives.
No effective way to prioritize or rank merchants for collections
Limited ability to determine the best rates and fees for a merchan

Solution

Complete view of merchant activity
Improved collections through statistical models to rate accounts by how likely they are to pay.
New pricing strategies — Analysts perform high-scale analysis to assess the impact of different fee structures on customer retention.

The impact

The project

Improved collection performance by filtering uncollectible accounts to 3rd parties and prioritizing highly collectible, high-value accounts. This generated millions of additional collections dollars per year projected due to the improved prioritization.
The repricing model facilitated repricing exercises for merchants which increased retention and generated additional millions above targeted revenue.

Integrated the Customer Relationship Management (CRM) system and other data sources to create a single source of truth: the Foundry Ontology, to understand Merchant activity specific to the chosen segment. Objects and relations were created for customers, their activity, pricing, payment terminals, billing, fraud, credit history, and more.
A business intelligence tool, Contour, and additional analysis tooling part of the Foundry platform perform analyses to determine new collections and pricing models. Scenario analysis of different pricing options is straightforward in Contour and a low code solution part of Foundry called Code Workbook.
Machine Learning models for prioritising collections and determining to reprice were implemented in PySpark codebook called Foundry Code repositories. Foundry’s Machine learning capabilities were used in a later stage for further analysis.

Improved collection performance by filtering uncollectible accounts to 3rd parties and prioritizing highly collectible, high-value accounts.
The repricing model facilitated repricing exercises for merchants which increased retention.

The project

Integrated the Customer Relationship Management (CRM) system and other data sources to create a single source of truth.
A business intelligence tool, Contour, and additional analysis tooling part of the Foundry platform perform analyses to determine new collections and pricing models.
Machine Learning models for prioritising collections and determining to reprice were implemented in PySpark codebook called Foundry Code repositories.