Key takeaways
- Subprime borrowers represent approximately 17% of active auto loan accounts but account for nearly two thirds of all delinquent loans — a concentration that demands a fundamentally different collections approach
- The 2022 to 2024 vintage subprime borrower entered a structurally different affordability environment: higher vehicle prices, longer loan terms, and a narrower margin between monthly income and payment obligation
- Traditional volume-based collections on a high-volume subprime book can produce 10,000 dials, 800 connects, 80 promises to pay, and just 24 actual payments — a 0.24% conversion rate on dial volume
- Contact rate and promise-to-pay volume are input metrics; cure rate and dollars per agent hour are the outcome metrics that actually predict portfolio recovery
- Rebuilding collections strategy around daily AI scoring on intent to pay, channel affinity, and affordability produced a 66% lift in collections per agent per hour for one subprime auto lender
Subprime borrowers hold approximately 17% of active auto loan accounts. They account for nearly two thirds of all delinquent loans.
That concentration is not a credit underwriting problem at this point. It is a collections problem specific to this segment. The strategy most lenders are running was not built for this borrower, in this environment, at this scale.
The stress is not evenly distributed. It is concentrated precisely in the segment where collections strategy tends to be the least differentiated.
The 2022 to 2024 vintage borrower is not who your collections team thinks they are
Every collections team working this portfolio has a mental model of who their borrower is. That model was largely built on the borrowers from prior cycles, specifically 2015 to 2019 vintage accounts where the affordability math was different.
The borrower who originated a loan in 2022, 2023, or 2024 entered a structurally different situation. Vehicle prices reached record highs, loan terms stretched to 72 and 84 months to keep monthly payments manageable.
Now, interest rates have climbed sharply. The margin between what the borrower earns and what the loan costs narrowed to a point where a single irregular paycheck, a delayed direct deposit or an unexpected expense, is enough to produce a 30-day flag.
These borrowers are not delinquent because the payment competes with rent and groceries in a budget that does not have room for all three. The collections conversation that worked on a 2018 vintage borrower, the tone, the urgency, the assumptions about what it takes to get a commitment, does not work the same way on this one.
Auto loan delinquency management built on prior-cycle assumptions is not just ineffective on this portfolio. It is actively making recovery worse by applying pressure at moments when the borrower needs a structured path, not an escalation.
What the standard approach actually produces
The mechanics of traditional subprime auto collections look rational until you measure outcomes. One lender's daily activity before working with Prodigal:
What the standard approach actually produces
10,000 dials. 24 payments.
What these numbers reveal
0.24%
Conversion rate from total dial volume to actual payment. This is the real return on the activity investment.
RPC ≠ resolution
Reaching the right party only means you reached them. It does not mean they will pay.
PTP ≠ payment
A promise to pay means someone said they would. 80 promises produced 24 payments.
The same investment in agent wages, directed by a strategy that optimizes for dollars collected per agent per hour, produced a 66% lift in collections performance for this lender.
Optimizing for connects and promises is optimizing for activity, not for dollars collected. The two are related but they are not the same, and in a high-volume subprime operation the gap between them is where recovery performance is lost.
The metric trap in auto loan delinquency management
Contact rate and promise-to-pay volume are the two metrics that dominate subprime auto collections reporting. Both are inputs, neither is an outcome.
Contact rate measures how often you reached someone. Promise-to-pay volume measures how often someone agreed to pay. What both miss is whether the account actually resolved.
Cure rate is what predicts portfolio recovery, it measures the percentage of delinquent accounts that return to current status. Cost per cure measures what it took to get there, and dollars per agent per hour measures whether the labor investment is generating proportional returns.
On a 2022 to 2024 vintage subprime portfolio, calling the same borrower three times a day does not change the fact that their paycheck has not arrived yet. Activity volume does not cure accounts.
Effective auto loan delinquency management starts by measuring what actually does.
The metric trap
Input metrics vs outcome metrics in auto loan delinquency management
Outcome metric
Cure Rate and Dollars Per Agent Hour
What actually matters
Cure rate measures accounts that returned to current status. Dollars per agent hour measures whether the labor investment is generating proportional returns.
These are what predict portfolio recovery. A 66% lift in collections per agent per hour came from building strategy around these instead.
What a segment-aware subprime auto collections strategy looks like
The alternative to volume-based collections is not fewer contacts. It is smarter ones.
A subprime auto lender working with Prodigal rebuilt their collections strategy on daily AI scoring across three variables: intent to pay, channel affinity, and affordability.
Every account received a score each day, and outreach was timed to when payment likelihood was highest, not to a static call schedule.
- First-time delinquents, habitual late payers, and borrowers in genuine hardship require completely different approaches.
- A first-time delinquent needs a clear, low-friction path to resolution.
- A habitual late payer needs a different cadence and channel.
- A borrower in hardship needs structured options: an extension or a payment arrangement.
Collapsing these into a single treatment workflow produces the worst outcome for every segment.
Segment-aware collections
Three borrower profiles. Three different approaches.
Collapsing these into a single treatment workflow produces the worst outcome for every segment.
Barrier to payment
Timing, not willingness. Paycheck delayed or budget temporarily disrupted.
What they need
A clear, low-friction path to resolution with an accurate cure calculation. Not pressure.
Barrier to payment
Prioritization, not inability. This payment competes with other obligations.
What they need
Different cadence, different tone, different channel. Generic urgency creates avoidance, not action.
Barrier to payment
Capacity. The budget does not have room for the full payment right now.
What they need
Structured options — an extension or a payment arrangement — presented clearly and early.
Prodigal deployed 200+ persona-matched templates across these profiles with daily-updated scoring on intent to pay, channel affinity, and affordability. Result: 78% increase in digital engagement, 27% lift in the 30-day past due bucket, 45% increase in self-serve payments.
Frequently asked questions
Frequently asked questions
Subprime auto collections strategy: common questions
How quickly do AI-driven collections models improve after deployment on a new portfolio?
Models begin learning from the first contact attempt. Every outcome — payment made, call ignored, message opened, promise kept or broken — feeds back into the scoring that determines the next outreach decision. Daily model updates mean the strategy adapts to your portfolio's specific borrower behavior within weeks, not months. The improvement compounds: a model that has observed 90 days of contact outcomes on your book is meaningfully more accurate than one operating on day one. This is why results tend to improve month over month rather than plateau after initial configuration.
How does having a similar segment of consumers elsewhere in a vendor's portfolio improve outcomes on mine?
A vendor who has worked across portfolios with similar borrower profiles — same delinquency stages, similar income profiles, comparable vehicle types — arrives with pre-trained intuitions about what works. They have already seen how a first-time delinquent on a 72-month subprime note responds to different outreach approaches, and how that differs from a habitual late payer. That prior learning accelerates time-to-value on your portfolio because the model is not starting from scratch. When evaluating vendors, ask specifically how many accounts with a similar subprime composition they have processed — and whether those portfolios are in production today, not archived from a prior engagement.
What does an AI-driven collections strategy actually measure, and how do I know it is working?
The metrics that matter are cure rate, dollars collected per agent hour, cost per cure, and self-serve payment rate. Activity metrics — contact rate, right party contacts, promises to pay — tell you what happened in the contact attempt. Outcome metrics tell you whether it produced a recovery. A well-structured deployment shows these side by side, with the AI-managed portion of the portfolio compared directly against the legacy process at each stage. Dashboards should give leadership visibility into contact attempts, payment rates, liquidation rate, AI containment rate, and dollars collected per agent hour — with drill-down by channel, delinquency stage, and borrower segment.
Can proCollect and proAgent be deployed independently or do they work better together?
Both can be deployed independently. proCollect optimizes your digital outreach strategy — channel mix, contact timing, message personalization, dialer list sequencing — and works alongside your existing human agent operation and current dialer and CRM infrastructure. proAgent automates inbound and outbound conversations across voice, SMS, and email. When used together, proCollect's daily scoring feeds proAgent's contact decisions, which produces compounding improvement: the right borrower reaches the right channel at the right time and receives an AI-handled conversation built on that intelligence.
What reporting is available for compliance audits and performance reviews?
Prodigal provides full audit-grade documentation including complete call logs with timestamps and disposition, consent verification records tied to each contact attempt, 100% call recordings and transcripts, interaction-level compliance and quality scorecards, and opt-out audit trails with timestamp and channel. Performance dashboards give leadership a single view of program health with drill-down by channel, delinquency stage, agent, or segment. Core metrics include contact attempts, connect rate, RPC rate, promise-to-pay rate, PTP kept rate, total dollars collected, liquidation rate, AI containment rate, and dollars collected per agent hour. Dashboards are web-based with role-based access, scheduled email summaries, and export capability.
.svg){kind=icon}
.png)
.webp)
.webp)
.jpg)
-min.avif)
.avif)
.avif)
.avif)
.avif)
.avif)