Collections Modelling
We build the models that predict which customers in arrears will pay, which will self-cure and which require support.
Collections
Customers in arrears: who will recover, who needs support, and how you reach them.
Collections Deliverables
Banking Case Studies
Banking
Independent validation of a Pillar 2A concentration add-on
A European bank needed to know that the capital it holds against concentrated lending would stand up to scrutiny, so it asked us for an independent view of the model behind the number.
Banking
Redeveloping every IFRS 9 model across two portfolios
A UK lender needed every model in its IFRS 9 expected credit loss suite rebuilt, at the point when the team that had built them was no longer there.
Frequently Asked Questions
-
Collections models are judged on net recovery after cost and after conduct, so ask a bidder how their segmentation changes what your agents do on Monday. At Gini we build the segmentation, the treatment path optimisation that assigns each account to the action with the best expected outcome, and the vulnerability identification that governs which treatments are available at all.
-
Gini treats the conduct constraint as part of the objective rather than a filter applied afterwards. Vulnerability indicators are built from the data you already hold and the treatments available to a flagged customer are restricted before optimisation runs, so the model cannot recommend an action you would not defend. Outcomes are then monitored by customer group, which is the evidence a supervisor asks for.
-
Yes. At Gini we measure what each existing path actually recovers net of cost and contact effort, which is often the first time those numbers sit side by side, then reassign accounts where the expected outcome is clearly better elsewhere. Paths that recover little and cost much are the usual finding, and retiring them frees capacity without any new model going live.
-
Often, yes. Payment behaviour, contact history and account activity carry signals of difficulty well before a customer discloses it, and those signals can be turned into a flag your agents act on. Gini is explicit about what the model cannot see, because a vulnerability framework that overstates its coverage is worse than one that names its limits.
