A consumer installment lender with a branch network across the southern United States runs Vector on early-delinquency accounts. In the early stages, the lender took a snapshot of how Vector performs against the benchmarks it tracks for its own collections team: how often a call reaches the right borrower, how often that borrower commits to pay, and what each contact costs. Vector reached the right borrower 40% more often per attempt, and nearly half of them promised to pay.
The short version
- Scope: 13,609 accounts in early delinquency, 21,988 outbound calls over one week.
- Right-party contact: 4.9% of attempts for Vector against the lender's 3.5% human benchmark.
- Outcome: 49% of right-party contacts ended in a promise to pay
- Cost: $12 per right-party contact, against a fully loaded $82 for a human collector.
- Every call, consent event and disclosure was logged.
01. The set-up
The lender's branch staff double as its collections team, so capacity, not willingness, limits how many delinquent accounts the branches can work each week. The question for the pilot was simple: on a delinquency bucket the branches already handle, could Vector reach more borrowers and secure more commitments at a lower cost per contact, inside the same rules?
The lender set the rules. Calling windows in the borrower's local time, contact frequency limits, consent status, the exact wording of required disclosures, and the point at which a conversation had to go to a person (any hardship signal, any request for a settlement or payment plan).
Vector was given the accounts as one mission, with a single outcome to work toward on each, and measured against the benchmark the lender already tracks for its own collectors.
02. The measurements
Right-party contact
Right-party contact is the share of attempts that reach the actual borrower.
Promise to pay
Promise to pay is the share of those contacts that end with a commitment to a date and an amount.
Cost per right-party contact
Cost per right-party contact is what the lender spends to get one live conversation with the right person.
On contact rate and cost, Vector is ahead.
+40% right-party contact rate · 49% of contacts promised to pay · $12 per right-party contact
A human collector works one account at a time and stops at the end of the shift. During the pilot, Vector retried failed attempts the same day, moved between calls and text, and kept working accounts over days until it reached the borrower or hit the lender's contact limit. Reaching the right borrower 40% more often, at the same promise-to-pay rate, is what turned into almost $1m of commitments over four weeks.
03. What does the lender get beyond the number?
A record of every action. Each call is tied to the account, the rules in force at the time, the disclosures delivered, the borrower's consent status, and, where a decision went to a person, who approved it. Before the pilot, that record was assembled by hand for audits and disputes.
The collectors don’t go away. With Vector covering the early bucket, branch staff spend their time on the accounts that need judgment: hardship cases, disputes, and the later-stage balances where a conversation with someone who knows the borrower changes the outcome.
Common questions
Why is the lender not named?
Regulated lenders are usually cautious about public statements on collections performance, and we would rather publish verified numbers without a name than a name without numbers.
Is a 3.5% benchmark comparable to Vector's 4.9%?
Both figures are right-party contacts as a share of attempts, on the same delinquency bucket, under the same contact rules, measured by the lender's own reporting. That is the test we would apply to any pilot claim, ours included.
What happened on the calls Vector could not handle?
They went to a person. For this lender, Vector was tuned to escalate on any sign of hardship or any settlement request, which means some transfers were unnecessary. Vector always works within the lender's rules.
Start with one mission. Rather than trying to onboard yet another AI company that promises the world, start with one use case, one mission. Book a demo at vector.synthflow.ai.
.avif)
