A customer segmentation approach in commercial banks

Abstract

The results of various recent analyses of commercial banking trends show that proper classification of borrowers is fundamental for the development of a successful business. The increasing competition in both banking system and non-bank institutions requires the use of modern market strategies and individual customer approaches. One of the main priorities in the banking sector is to improve customer segmentation and take it into account in the design and distribution of new products. A common tool to improve competitiveness is the designing of a special range of products and services targeting loyal customers or offering them special discounts for existing products. This represents the so-called “loyalty program”, which includes the issuance of various types of cards for such customers. Three clusters (segments) of loyal borrowers: “platinum,” “gold,” and “silver,” are identified in the present work, using K-means clustering. A database of 100 borrowers from a commercial bank branch that took secured consumer loans is analyzed. The clients are defined as loyal based on their credit history (they have less than 3 missed payments for the last year). Three variables are used for their segmentation. Initially, the initial segmentation variables are taken as input data for the analysis, and further study on standardized segmentation variables is carried out. The potential segmentation strategies are formulated depending on the leading segmentation variable. A comparative analysis of the results of both methods examined (with initial and standardized segmentation variables) and of а two-step clustering (obtained in a previous study from one of the authors) is made within each of the strategies. It is specified which type of cluster analysis suits best to each of the strategies.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 31 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.