2021 · Machine Learning

PRISM — Pricing & Wallet Share Optimisation

Clustering, benchmarking and scoring to make pricing, utilization and wallet-share decisions more scientific.

15% income growthHighlighted outcome / scale
  • ML clustering created comparable peer groups
  • Decile benchmarking for pricing, wallet share, utilization & RAROC
  • 15% income growth in year one
Case study infographic
PRISM — Pricing & Wallet Share Optimisation infographic
Project detail

From business problem to measurable impact.

01

Business Problem

Relationship Managers could see their own portfolios, but had limited bank-wide comparability when asked to improve pricing, utilization or wallet share. Because every corporate client could be described as “unique,” target discussions were often subjective and difficult for management to defend consistently.

02

My Role

Helped combine business context, data science and corporate-finance logic into a decision system: defining peer-group factors and commercial metrics, translating benchmark positions into target settings and supporting leadership/RM adoption and monitoring.

03

Solution

PRISM clustered relatively homogeneous corporates using factors such as credit rating, industry, ownership, size and leverage. Within each peer group, customers were benchmarked using deciles on interest rate, collection and loan wallet share, utilization and RAROC; business leadership then selected controlled target movements and priorities.

04

Technology & Methods

ML clustering · benchmarking · decile scoring · RBI total-borrowing data · pricing analytics · wallet-share metrics · utilization · RAROC · dashboards and monitoring

05

Leadership Lesson

“Analytics can make commercial judgment more scientific without removing human judgment: peer-based evidence narrows the debate, while business leaders retain control of targets and priorities.”

06

Detailed Case Study

The infographic above is the executive summary. The full portfolio document contains the methodology, operating model, governance and implementation narrative at an externally shareable level.

View detailed PDF ↗

Confidentiality note. Customer-level data, proprietary model details, source code and confidential implementation information are intentionally excluded.