Cx Collections Conduct

Collections and financial-difficulty conduct

These are the highest-stakes conversations a support operation has. The customer is under financial pressure, the agent has a collections objective, and the outcome affects whether someone can pay for essentials. In many sectors there are specific duties about how this is handled.

The structural problem is simple and worth stating plainly: the agent has an objective that can conflict with the customer's interest. Every finding below traces back to that, and the incentive structure is usually the root cause rather than the individuals.

Requirements vary by jurisdiction, product and circumstance and are a compliance and legal determination. Nothing here states one.

What to audit

Recognition of difficulty. Did the agent identify that the customer was in financial difficulty, as opposed to simply late? These need different handling and the distinction is frequently missed. Signals: statements about affordability, prioritising essentials, other debts, a change in income, or asking for time.

Forbearance offered, and appropriate. Where a difficulty is identified and options exist — a payment plan, a hold, reduced payments, a breathing space — were they offered proactively, or only if the customer knew to ask? Options offered only on request are options most customers never receive, and that pattern systematically disadvantages the least confident. Measure the proactive-offer rate.

Affordability actually assessed. Where a payment plan was agreed, was the amount based on what the customer said they could afford, or on what the balance divided by twelve came to? A plan the customer cannot sustain fails, generates another arrears event, and leaves them worse off — check plan breakage rate as an outcome measure.

Pressure absent. The core conduct test:

• Urgency that is not real, or consequences implied that do not follow.
• Repeated contact beyond what is reasonable, or after the customer engaged.
• Discouraging a customer from a course they are entitled to take, including complaining, seeking free debt advice, or exercising a right.
• Asking a customer to prioritise this debt over essentials or over a priority debt.
• Continuing to press after a customer has disclosed vulnerability or distress.

Vulnerability recognised. These conversations carry a high concentration of vulnerability signals, and recognition and adaptation should be measured — recognising it and then continuing identically is a finding.

Referral to independent debt advice, where that is expected, and whether it was framed as genuinely useful rather than as a formality.

Contact conduct — frequency, timing, channel, and whether the customer's stated preferences and any restrictions were respected.

Check the incentives first

Before reading a single conversation, look at what the collections function is measured and paid on. Where the objective is a collection rate, a promise-to-pay rate, or a right-party- contact target, expect pressure and under-recognition of difficulty — reliably, from ordinary people, because that is what was asked for.

If the audit finds a pressure pattern and the incentive rewards it, the incentive is the finding, and reporting the agents instead would be both unfair and ineffective. Look also at the script and the call flow: pressure is frequently written down as the recommended approach.

Outcome measures beyond the conversation

Conversation review tells you what was said. These tell you whether it worked:

• Plan breakage rate, by how the plan amount was set. High breakage on plans set without an affordability discussion is direct evidence of unaffordable plans.
• Repeat arrears after a resolution.
• Complaint rate on collections contacts, and their themes.
• Outcome variation by market, channel, team and agent for comparable circumstances.
• Whether customers who disclosed difficulty ended up better or worse off than comparable customers who did not disclose. If disclosure makes things worse, that is the headline finding and it outranks everything else, because it teaches customers not to disclose.

Sampling

• Risk-weight toward first-arrears conversations, cases with vulnerability signals, plans that later broke, and complaints.
• Keep a random core, because you will be asked for a population rate and only the random core supports one. Never pool them.
• Voice is the dominant channel here and its conduct signals — tone, interruption, pace — do not survive transcription. Verify a sample against audio before concluding anything about pressure, and state the limit for the rest.
• Cover every language.

Guardrails

• Do not determine whether a duty was breached or whether remediation is required. Compliance and legal decide, and there may be a notification clock.
• Do not report individual agents without checking the incentive, script and call flow. A pattern spanning several people is systemic.
• Vulnerability findings stay aggregated, with no named lists and no disclosure text.
• Do not use the audit to improve collection rates. If that becomes the objective, the findings will be read for how to press more effectively, which inverts the purpose. Say so if it happens.
• A pattern of pressure on vulnerable customers is an escalation, not a report line.
• Cite ids; quote only the specific exchange where the wording is the finding. These files contain financial circumstances, health disclosures and vulnerability information, and the report will circulate to compliance and possibly externally.

Present results to the user

1. The incentive structure, and which of the observed behaviours it would predict. First, because it frames everything.
2. Recognition of difficulty, as a rate, distinguished from simple lateness.
3. Proactive forbearance-offer rate, separately from offered-on-request.
4. Affordability assessment, and plan breakage by how the amount was set.
5. Pressure findings, by category, with the voice-verification caveat stated.
6. Vulnerability recognition and adaptation, aggregated.
7. The disclosure test — whether customers who disclosed difficulty fared better or worse. If worse, lead with it.
8. Outcome variation across markets, channels, teams and comparable circumstances.
9. What the sample supports — risk-weighted versus random core, and which population each result describes.
10. What needs a compliance determination, and anything escalated immediately.