L'entreprise
Volume 15, Numéro 1, Pages 69-82
2026-09-08
Authors : Djaballah Noureddine .
Summary: Customer attrition is a strategic concern for banks, particularly in the corporate segment where the weakening of a relationship is rarely sudden. A company may keep its account open while gradually redirecting its flows, financing needs and day-to-day banking activity to another institution. For this reason, corporate churn must be understood not only as account closure, but also as a progressive decline in engagement. This paper examines how data can help Société Générale Algeria (SGA) better understand, monitor and anticipate attrition among corporate clients. The study combines three complementary sources of evidence: semi-structured interviews with internal stakeholders from business and data functions , a survey of inactive corporate clients, and a prototype churn-prediction model developed in Python using a Random Forest algorithm. The survey focused on inactive corporate clients in March 2025 (N = 250), defined as clients with no transactions between January and March 2025 while remaining active in March 2024. A quota sample of 150 clients was selected , and 100 usable responses were collected, giving a response rate of 67%. The results point to weak banking engagement: 78% of respondents do not consider SGA their main bank, and 62% only hold a business current account without credit facilities. The main reasons for inactivity are largely within the bank's sphere of influence: service quality, relationship management, communication and pricing . The statistical analysis shows that relationship satisfaction, lead-time satisfaction and offer-related satisfaction significantly influence overall satisfaction, while digital satisfaction is not significant in the model. The logistic regression displays strong explanatory power (Nagelkerke R² ≈ 0.885). The Random Forest prototype (2,673 clients, 35 variables) achieves very high test performance (accuracy ≈ 0.99; AUC ≈ 1.00). These results suggest strong operational potential for early detection, although time-based validation and robust leakage controls are necessary before production deployment. From a managerial perspective, the study recommends a proactive attrition-management framework that combines organizational action, especially a dedicated corporate CRCM team, operational improvements in lead times, complaints and the credit journey, and data-driven steering through segmentation, scoring and retention playbooks (Buttle, 2004; Durand-Megret & Van Laethem, 2022).
Customer ; attrition ; data-driven marketing ; banking corporat clients ; retention ; Random forest
Datoussaid Aimad
.
Labiad Naima
.
pages 1413-1432.
Lanseur Akila
.
Boukellal Sabrina
.
pages 852-867.
Adnane Amina
.
pages 1163-1174.
Belhadi Abdelkader
.
Nezai Azzeddine
.
pages 501-517.
Idiri Yanis
.
Belkhiri Aimad Edine
.
Azazga Imene
.
pages 735-749.