Job family

Assessing collections and retention

The conversations nobody volunteers for — asking for money that is overdue, or keeping a customer who has already decided to leave.

Collections and retention are usually managed as separate departments, but they are the same conversation with different stakes. In both, the customer wants the call to end, the agent is paid to keep it going, and the rules about how far they may go are written down somewhere the agent has to remember mid-sentence. That combination — adversarial contact, explicit conduct rules, and a bonus attached to the outcome — is what makes this the family where screening most obviously fails, because the standard screen is a compliance module and a multiple-choice quiz. A quiz establishes that the candidate knows the rule. It says nothing about whether they follow it at minute nine of a call with someone who is crying, or lying, or shouting, when following it costs them the outcome.

The CFPB's complaint data is effectively a published list of the behaviours a simulation should be watching for, which is unusual and useful. Among consumers who complained that a debt was not owed, 60 percent said the debt was not theirs and 28 percent reported identity theft — that is, the largest single category of complaint is a collector continuing after the consumer disputed the debt. Among complaints about communication tactics specifically, 51 percent concerned frequent or repeated calls and 34 percent concerned continued contact after the consumer asked for it to stop. Both are directly observable in a transcript. A candidate who keeps pressing after a dispute, or who acknowledges a stop request and then asks one more question, has demonstrated the exact behaviour that generates the complaint volume — and has done it in a simulation, at no cost to a real consumer.

The commercial skill is separable from the compliance one, and good designs score both. In collections, the outcome that matters is not a promise to pay but a payment arrangement the person can actually keep: an agent who extracts an unrealistic commitment books a win this week and creates a broken arrangement, another contact cycle and a worse recovery next month. In retention, the equivalent is a save that holds — an offer matched to why the customer is actually leaving, rather than a discount reflexively thrown at someone who was going anyway, or a retention promise that quietly becomes a mis-selling problem. Both also require recognising vulnerability and changing approach, which the better-run teams already train and none of them screen for. Two design constraints worth stating to any buyer: score whether the stop request was honoured and whether the arrangement was affordable, never the candidate's accent or manner; and never build a scenario that rewards inferring anything about the customer's personal circumstances beyond what they say on the call.

Why this work can be assessed

Both are single adversarial calls governed by explicit rules, and compliance is visible line by line in the transcript. A simulation can put the candidate at the exact fork where the profitable move and the permitted move come apart.

Roles in this family

Sources

Every figure on this page is traceable. Where a claim could not be sourced it is stated qualitatively instead.

1. US Bureau of Labor Statistics, Occupational Outlook Handbook, Bill and Account Collectors, 2025, https://www.bls.gov/ooh/office-and-administrative-support/bill-and-account-collectors.htm
2. Consumer Financial Protection Bureau, Fair Debt Collection Practices Act CFPB Annual Report 2025, November 2025, https://files.consumerfinance.gov/f/documents/cfpb_fdcpa-2025-annual-report_2025-11.pdf