Success stories



D3S team developed algorithms to predict overdue payments from past customer behavior, including peer comparison (country, sector, size, time period, etc). The model was packaged and pluged to SaP to deliver near real time alerts. It is used by 30 cash collector in several countries.


We started the project with transactional data over the last 3 years, extracted from the order-to-cash process and CRM systems. It represented more than 1 million invoices to analyze, with various complexities in clearing mechanisms. All data were collected in .csv format and uploaded to a PostgreSQL database before processing.

Our Tech Stack : Python, SKL, PostreSQL and near real time API to SaP through BW.


A CAC40 industrial company, with more than 100 000 customers is facing overdue. In the past years, a efficient collection process was implemented with more than 30 FTE calling customers to secure on-time payments. The head of credit management was willing to leverage available data and go one step further.

We conducted 18 interviews with referees all across the globe, to understand the data structure and collection processes. Data scientists developed complex SQL queries to prepare features, and the algorithm was designed to automatically learn customers habits in a specific context (customer sector, size, payment history, FX rates, holidays, invoicing pattern, etc).

The proposed solution was validated for GoLive by Top executives and pluged to SaP.


+20% productivity gains were demonstrated, when compared to the current methodology to prioritize collection efforts. The model can predict overdue invoices at the posting date, to prioritize calls prioritized based on “best expectation of gain” taking into account past behaviors of each customer.

Feedback from cash collection teams was excellent, with less time spend on low value-added tasks and a stronger hit rate on pre-dunning actions. Industrial solution was fully transferred to the IT department and integrated to SaP.