In-house, agency or hybrid: choosing a collections model

Every lender past a certain size makes this choice, revisits it after an incident, and often lands somewhere in the middle. The three models are not ranked. They trade different things against each other, and which trade is right depends on the book, the stage of the lender and how much conduct risk sits with the brand.

In-house

Best where conduct risk is high, where the borrower relationship continues after the arrears are cleared, and where the early bucket is large. You control training, tone and escalation directly, the data never leaves, and the feedback loop into underwriting is short. The costs are fixed, hiring is continuous, and geographic coverage is expensive to build for a book that is spread thin.

Agency

Best for reach, for deep buckets where the economics only work on commission, and for surge capacity. You buy coverage in cities where you have no presence and specialists in situations you handle rarely. What you give up is visibility and, unless the contract is unusually tight, control of the conversation. Complaints still arrive at your brand, and the regulator still regards the borrower as yours.

Hybrid, which is what most large books end up doing

• In-house for pre-due through the early bucket, where tone matters most and volume is highest.
• Agency for the deep buckets and for cities without a stationed team.
• A single platform across both, so an account handed to an agency does not disappear from view.
• One conduct standard, audited the same way regardless of who is holding the phone.

The questions that decide it

• If a borrower complains about an agency caller next month, can you produce the recording within the day?
• When an account is recalled from an agency, what happens to the copy of the data they hold?
• Do agency users log in individually, so an access log names a person?
• Can you compare agency performance to in-house on the same cohort, or only on the accounts each happened to receive?

That last one is where most comparisons fail. Agencies usually get the harder accounts, then get judged against in-house numbers built on easier ones. Unless you can hold the cohort constant, the comparison is measuring allocation rather than performance.

The direction of travel

As a book grows, the early bucket grows fastest and is the cheapest to automate, which pushes it in-house. The deep tail grows slowly and stays specialist, which keeps it with agencies. Most operations that get this right end up doing more of their own early work over time and less of their own late work, rather than choosing one model for everything.