The Path That Doesn’t Require Talking: Designing Self-Service Collections | HL Hunt
The Path That Doesn't Require Talking: Designing Self-Service Collections
A substantial share of people who fully intend to resolve an account will not have a conversation about it. Not because they refuse to pay — because the conversation means explaining a personal situation to a stranger who is assessing it. An operation whose only genuine route to resolution runs through an agent loses those people entirely. And the usual response, a payment portal, mostly doesn't fix it: a portal that only takes payment in full serves the people who could have paid anyway, which is the group that needed a channel least.
What you'll learn
Who this channel is for
Not a cheaper version of the phone. A route for people the phone doesn't reach.
| Group | Why the phone doesn't work |
|---|---|
| Won't discuss it | The conversation itself is the obstacle |
| Can't call in business hours | Per our time analysis, shift work |
| Prefers digital for everything | Generational and habitual |
| Needs time to think | A call demands an immediate answer |
| Language or communication needs | Text is easier than speech |
Row four is underrated. A call requires deciding what you can afford while someone waits — and a portal lets someone look at the numbers, work it out, and come back, which produces more realistic arrangements and fewer that break.
Per our time analysis, row two is a structural exclusion rather than a preference. An operation available only in business hours has made itself unavailable to the people whose hours are least their own.
Why the usual portal fails
The specific design failure, and it's almost universal.
A typical portal offers: view balance, pay in full, pay an amount, call us.
Consider who arrives:
- Someone who can pay in full — served, and they'd have resolved anyway.
- Someone who can pay part — makes a payment with no agreement about the rest, so the account stays delinquent and treatment continues.
- Someone who needs an arrangement — hits "call us", which is the thing they came here to avoid.
- Someone who disputes it — no path.
- Someone in difficulty — no path.
Three of five have nowhere to go, and they're the three who most needed the channel.
And the second case is actively harmful. A part payment with no agreement leaves the person believing they've engaged and the operation still treating them as delinquent — which produces a contact that feels like bad faith to someone who just paid.
Authority is the whole thing
The design principle, and it's the same finding as our place analysis in a different channel.
A path that ends in a referral is not a path. What a working self-service route must be able to do:
- Accept an arrangement within defined bands, without a human approving it.
- Accept a settlement offer within bands, per our settlement analysis.
- Change a payment date, which per our timing analysis resolves a whole category of accounts.
- Pause treatment while an arrangement is in place, which has to happen automatically.
- Confirm in writing, immediately.
- Handle a missed instalment without dropping the person back to the start.
Item one is the difference between a portal and a channel. Per our inbound analysis, delegated authority at the point of contact is what converts a contact into a resolution — and it applies identically here.
And item four is where implementations break. An arrangement agreed digitally that doesn't stop the dialer is worse than no arrangement, because the person did what was asked and got called anyway — which per our restrictions analysis is the same propagation failure in a different guise.
The arrangement that doesn't stop the calls
Someone sets up a plan online at ten at night and gets a collection call the next morning because the arrangement hadn't propagated. That single failure destroys the channel's credibility and teaches the person that the digital route doesn't count.
Authentication
Where most abandonment happens, and it's invisible because it happens before anything is recorded.
Someone who can't get in leaves no record of what they were trying to do. Common causes:
- Requiring a reference number from a letter they don't have.
- Requiring registration before anything can be seen.
- Data that doesn't match what the person would type.
- A one-time code to a number that's out of date — the exact case where contact data is stale.
- Too many fields.
What to do:
- Use a unique link from the outreach that pre-identifies the account.
- Accept multiple identifiers, so a missing reference number isn't fatal.
- Allow progress before full verification where appropriate — seeing that an account exists is different from seeing its details, and that distinction is worth designing around.
- Provide a recovery path for people who can't authenticate, which is a defined route rather than a dead end.
- Test it on a phone, which is where most people arrive.
This has to be balanced against genuine security requirements, which matter in this context more than most — disclosing account information to the wrong person is a serious failure. The goal is removing friction that isn't doing security work, not weakening verification.
Showing the position
What the person needs before they can decide anything.
- What's owed, in a figure that matches what they'll see elsewhere.
- What it's made of — principal, interest, charges — since per our composition analysis a balance that doesn't reconcile to the original is frequently why someone disengaged.
- Every account you hold, not one at a time.
- What happens next if nothing changes, stated plainly and accurately.
- What options exist, with the arrangement option as prominent as full payment.
- What a plan would look like, calculated rather than described.
The second and fifth do the most work. A person who can see how a balance grew is more likely to engage than one facing an unexplained number — and an arrangement option presented as prominently as full payment tells them the thing they most need to know, which is that partial resolution is possible.
And per our framing analysis, the language matters. A page written in the register of a demand letter undoes the reason the person chose this channel, which was to avoid being spoken to that way.
What must reach a person
The boundary, and it needs deliberate detection rather than a contact link.
Per our disclosure analysis, some situations must not be handled by a form:
- Any indication of difficult circumstances, including in free-text fields.
- A dispute about whether the amount is owed, per our dispute analysis — which must be recognized as a dispute however it's phrased.
- Notification of a bereavement, per our deceased accounts guide.
- Mention of legal representation or insolvency.
- Anything indicating the person needs support.
What this requires:
- Free-text fields that are actually read — and monitored, not just stored.
- Detection of relevant language, with escalation.
- A visible route to a person, offered rather than hidden.
- Suspension of automated treatment on any such signal, pending review.
- Handling outside hours, since that's when people use this.
Item four is the one that matters most. Someone who discloses a difficult situation at eleven at night and then receives an automated reminder the next morning has been failed by the design — so the suspension has to be immediate and automatic rather than waiting for someone to read the message.
Measuring abandonment
The measurement approach, and it's different from how these are usually assessed.
Completions tell you how many got through and nothing about where everyone else stopped. What to instrument:
- Arrivals, by source.
- Authentication attempts and failures — frequently the largest drop and the most overlooked.
- Balance viewed, without further action.
- Options viewed, without selection.
- Arrangement started and abandoned, at which step.
- Completions, by type.
- Arrangements that hold, per our plan analysis.
The gap between "options viewed" and "arrangement started" is the most informative number available — it's people who engaged, saw what was offered, and decided none of it worked. That's a product finding rather than a conversion problem.
And per our segmentation analysis, use a holdout to establish that the channel is adding resolution rather than moving it — some people using a portal would have called. The holdout separates genuine incremental recovery from substitution.
Getting people there
The channel only works if people arrive, and arrival is mostly a design problem.
- A direct link in every communication, prominent — per our letters analysis, at least as prominent as the phone number.
- Unique links that pre-identify, removing the authentication barrier.
- An offer of the digital route on calls, particularly where someone declines to discuss it.
- Text as an entry point, subject to consent requirements.
- Timing that suits the recipient rather than the operation — evenings and weekends produce the engagement that business hours can't.
- An honest description of what the person can do there, since "make a payment" and "set up a plan" attract different people.
The last is the cheapest fix available. A message saying "pay your balance online" reaches only people who can pay their balance; one saying "set up a payment plan online, without calling" tells the people who came for that exact thing that it exists.
A path that actually ends somewhere
HL Hunt AI Debt Collection provides a self-service path with delegated arrangement authority within configured bands, immediate treatment suspension on agreement, hardship and dispute detection with routing, and step-level abandonment reporting.
Frequently asked questions
The conversation means explaining a personal situation to a stranger who's assessing it. Many people who fully intend to resolve will avoid that exchange.
The ability to agree an arrangement, not just pay in full. A pay-only portal serves people who'd have resolved anyway.
No. Hardship signals, disputes, and bereavement must reach a person, which requires deliberately detecting them rather than offering a contact link.
By abandonment at each step. Authentication is frequently the largest drop and leaves no record, since someone who never got in never said what they wanted.
Key takeaways
- This is a route for people the phone never reaches, not a cheaper version of a call.
- A pay-in-full portal leaves three of five arrivals with nowhere to go, and they're the three who needed it.
- A part payment with no agreement leaves the person feeling engaged and the system still treating them as delinquent.
- An arrangement that doesn't stop the dialer destroys the channel's credibility in one morning.
- The gap between options viewed and arrangement started is a product finding, not a conversion problem.
- Say "set up a payment plan without calling" rather than "pay online" — it's the cheapest fix available.
Reach the people who won't pick up
Get started with HL Hunt AI Debt Collection for digital arrangement setup with real authority, automatic treatment suspension, out-of-hours availability, and holdout measurement of incremental recovery.
This guide is educational and does not constitute legal or compliance advice. Requirements governing collection communications, consent for electronic and text contact, required disclosures, authentication and data protection standards, dispute handling, and the treatment of consumers in vulnerable circumstances vary by jurisdiction and apply to digital channels as fully as to any other. Consult qualified counsel and your compliance function before deploying a self-service path.