Retail Credit Limit Optimisation

Context

A major European fashion retailer with €500M in annual revenue offered customer credit as a standard part of its purchasing experience. Credit limits were assigned using static rule-based policies that treated customers as broadly similar — with no mechanism to account for individual purchase behaviour, repayment track record, or propensity to return items.

The Challenge

The credit limit policy was failing in two directions at once — and both had a real cost.

Customers assigned limits that were too high relative to their actual behaviour tended to over-order and return a large proportion of what they bought. Returns in fashion retail are not simply cancelled revenue — they generate active costs: processing, inspection, restocking, and in many cases markdown losses on items that cannot be resold at full price. Over-allocation was quietly driving losses that weren't immediately visible as a credit problem.

In the other direction, customers with genuine purchasing intent and low return rates were being constrained by limits too low to accommodate their basket size. These were the retailer's most valuable customers — and they were being held back.

Our Approach

The insight driving the model design was that credit limit optimisation in retail is not purely a risk problem — it requires optimising across three dimensions simultaneously: a customer's future purchase potential, their credit risk, and their likely return behaviour. A limit that looks profitable on purchase volume alone may not be once returns are factored in.

We built a model that incorporates all three dimensions to determine an optimal limit for each customer individually — one that maximises net revenue after returns rather than gross spend. Limits are assigned based on what each customer is likely to buy, keep, and pay for — not what segment they fall into.

Results