Collections KPIs for Auto Lenders: Definitions, Formulas, and the Public Benchmarks That Actually Exist

A collections program is run on about a dozen numbers, and most of them are computed differently from one lender to the next. Delinquency can be counted by balance or by account. A contact can mean a conversation with the borrower or any answered call. A charge-off can happen at 120 days or at 180. The numbers still get compared, in board decks, vendor reviews and trade articles, as if they were the same thing.

This guide gives each KPI a plain definition and a formula, lists the public benchmarks that actually exist for auto lending with the basis each one uses, and names the comparisons that mislead. It is written for the lender, dealer principal, collections manager or CFO who has to decide whether a number is good, and for anyone reading a report from an outsourced team. It is orientation, not legal advice; the classification of any collections engagement is a question for counsel.

What a collections KPI is, and why the denominator matters

A key performance indicator is a ratio: something counted in the numerator, something it is divided by in the denominator, over a stated period. Change any of the three and the number changes without anything on the floor having changed. Delinquency is the clearest case. The Federal Reserve Bank of New York reports that 5.49% of outstanding auto loan balances were 90 or more days delinquent in the second quarter of 2026. That is a balance-weighted stock: dollars in the bucket divided by dollars outstanding, at quarter end. Experian, as reported by Auto Remarketing, puts 30-day delinquency at 2.39% and 60-day delinquency at 0.90% for the same quarter, measured on accounts. A large balance that is 90 days late weighs more in the first number than in the second; a hundred small delinquent accounts weigh more in the second.

The period matters as much as the basis. The New York Fed's flow into serious delinquency, 3.00% for auto loans in Q2 2026, is defined in the release as an annualized share of balances that newly became at least 90 days late in the quarter, divided by balances that were current or less than 90 days past due in the previous quarter. A lender's own monthly roll into 90+ is the same idea over one month, not annualized. Multiplying a monthly rate by twelve gets close but not to the same number, because balances that roll early in the year are no longer in the denominator later in it.

Three questions therefore sit in front of every KPI in this guide: is the numerator counted in accounts or in dollars, what population is in the denominator, and over what period. Write the answers into the report header. A number without them is not a KPI; it is an anecdote.

The KPI table: definition, formula and what distorts it

The table covers the KPIs a collections program is run on, from the contact funnel through to loss. The formulas are the common operating forms; a portfolio may use a variant, but it should use the same variant every month. The right-hand column is the one to read first when a number looks better than it should.

Two rows carry an outside reference point. The attempts row is bounded by Regulation F: a debt collector is presumed to comply with the harassment prohibition when it places no more than seven calls about a particular debt within seven consecutive days and none within seven days after a telephone conversation about it, and is presumed to violate it when either frequency is exceeded. The dollars-per-hour row has a cost floor in the Bureau of Labor Statistics mean hourly wage of $23.59 for bill and account collectors in May 2025, a wage-only figure before payroll taxes, benefits, recruiting, training, seat cost and supervision.

Leading indicators and lagging ones

The KPIs sort into two groups by how fast they respond. Right-party-contact rate, promise-to-pay rate, promise-kept rate and the early roll rates respond within a week of a change in scripts, staffing or dialer settings. Cure rate follows within a month. Charge-off rate, net loss and recovery rate respond a quarter or more later, and they respond to underwriting, pricing and repossession policy as much as to collections. A floor managed on charge-off alone is managed on last quarter's decisions.

The market data shows the same split. The New York Fed reports both a stock, the 5.49% of auto balances 90 or more days late, and a flow, the 3.00% of balances newly entering serious delinquency, up from 2.93% a year earlier. The flow moved first; the stock, which stood at 5.60% in the first quarter of 2026, is where earlier flows accumulated. “Delinquency rates across most products have held steady over the past two years,” said Joelle Scally, Economic Policy Advisor at the New York Fed, in the Q2 2026 release. “Still, new delinquencies for auto loans and credit cards remain at elevated levels, a trend we'll continue to monitor.” The distinction she draws, a steady stock and an elevated flow, is the one a collections manager should draw every week.

In practice the funnel is the working model: attempts produce right-party contacts, contacts produce promises, promises are kept or broken, and kept promises cure accounts before they roll. Each stage has a rate. When cure rate falls, the funnel shows whether the cause is reachability (RPC down), conversation quality (PTP down) or affordability (kept rate down), and each has a different fix. The BHPH collections best practices guide sets out a contact cadence built around those stages.

The public benchmarks that actually exist

Public benchmarks for auto collections are few, and each measures a different population on a different basis. The table lists the ones that exist as of September 2026 with the institution, the period, the basis, and the comparison each can and cannot support. None is a target for a specific portfolio; they show the direction of the market and roughly where a prime, near-prime or subprime book sits relative to the whole.

One row needs a note. The Federal Reserve Board's charge-off and delinquency release is one of the few public sources that report a charge-off rate on consumer credit at a regular cadence, and it covers insured U.S. commercial banks only, and it covers insured U.S. commercial banks only. Its “other consumer loans” category, charged off at 1.21% and delinquent at 2.33% in the second quarter of 2026, includes auto loans held by banks alongside other installment credit, and excludes finance companies, captive lenders, credit unions and every buy-here-pay-here dealer. It is a reference for the banking sector, not a benchmark for a subprime portfolio.

What does not exist publicly matters as much. There is no public benchmark for right-party-contact rate, promise-to-pay rate, promise-kept rate or attempts per account on auto portfolios, because those are computed inside each lender's system on definitions that differ. A figure offered for any of them without an institution, a period and a denominator is a survey number at best. The same discipline applies to vendors: TEKS does not publish generic performance percentages, and the figures it reports to a client come out of that client's own system of record, on that client's definitions.

The comparisons that mislead

Most misreadings of collections numbers come from comparing two figures that share a name and nothing else. The list covers the ones that recur in board decks and vendor reviews.

• Balance-weighted against account-weighted. The New York Fed's 5.49% is dollars; Experian's 2.39% is accounts. A subprime book with small balances and a prime book with large ones can rank in opposite order on the two bases. Report both for your own portfolio and label each.
• Stock against flow. A 90+ share of balances is where past roll accumulated; a flow into 90+ is what is rolling now. Comparing this quarter's stock with last quarter's flow says nothing about whether collections improved.
• Annualized against monthly. The New York Fed annualizes its transition rates; most floors report roll rates per month. Twelve times a monthly roll is not the annual flow, because the population shrinks as accounts roll out of it.
• First-party denominators against third-party ones. A third-party agency's liquidation rate is cash collected divided by balances placed, on accounts that already defaulted and were often charged off. A first-party team's cure rate is on accounts that are late but live, with the relationship intact. The populations differ, and the difference is the one the statute turns on: the definition of a debt collector at 15 U.S.C. § 1692a(6) reaches those collecting debts owed or due another and excludes a creditor's own officers and employees collecting in the creditor's name. Where a particular outsourced program falls is determined with the client's counsel; the denominators should never be mixed regardless.
• Attempts against contacts. Attempts measure effort and compliance exposure; contacts measure output. Under Regulation F, more than seven calls about a particular debt in seven consecutive days creates a presumption of violation, so a rising attempts number is not evidence of a floor working harder in any useful sense. Report attempts as a ceiling check and contacts as the productivity figure.
• Bank rates against non-bank portfolios. The Federal Reserve's commercial-bank figures describe lenders that underwrite mostly prime borrowers. A near-prime or subprime book should expect materially higher delinquency and charge-off on the same definitions, and the gap is not a performance finding.
• Re-aged accounts against cured ones. An extension that moves an account from 60 days past due to current has not collected anything. Track cures by payment and cures by modification separately, or the cure rate will improve exactly when the portfolio's risk is rising.
• Complaint counts against conduct. The CFPB's guidance on its own Consumer Complaint Database is that a low number of complaints does not necessarily mean little or no consumer harm, that the database is not a statistical sample, and that volume should be read against company size and market share. A complaint rate is a signal to investigate, not a verdict either way.

How to read an outsourcing vendor's report

A report from an outsourced collections team should be readable against the same twelve KPIs, on the definitions the lender uses internally, and drawn from the lender's own system. If the vendor's numbers come from a separate dialer export the lender cannot query, the first question is not whether the numbers are good but whether they can be reconciled. TEKS agents work inside the client's CRM, DMS or loan-servicing platform for this reason: the report and the system of record are the same data.

Then draw the line between what a first-party support team can move and what it cannot. It can move right-party-contact rate, by working the numbers on file, verifying contact information and using the channels the customer answers. It can move promise-to-pay and promise-kept rates, through conversation quality and arrangements the customer can afford. It can move attempts per account within the Regulation F ceiling, note completeness, and the 30-to-60-day roll rate that follows from all of the above. It cannot move charge-off rate, net loss or recovery rate on its own, because those depend on who was approved, at what price, when the lender charges off and what a repossessed vehicle brings at auction. A vendor that promises a charge-off number is promising something outside its control.

The guides on when a dealership should outsource collections and how to choose a BPO partner cover the decision before the report; the first-party collections support page lists the operating standards TEKS runs to. Each month, ask the vendor for the items below.

• The definition of a contact, a promise and a kept promise, in writing, matching the lender's own.
• Both denominators for RPC rate: per attempt and per account attempted.
• Attempts per account per day against the seven-in-seven ceiling, with exceptions listed by account.
• Complaints by channel and disposition, including those the lender received directly, and the exceptions logged under Regulation F §1006.6, which presumes contact before 8:00 a.m. or after 9:00 p.m. at the consumer's location inconvenient, requires that a written cease request stop further communication, and requires an opt-out method in every electronic message.
• Dollars collected per collector hour, with autopay and self-cure dollars shown separately.
• Note completeness, sampled by a lender manager reading the notes rather than computed from disposition codes alone.

Reporting cadence

Cadence should follow how fast each KPI moves. Contact-side figures change daily and are reviewed daily during a launch; loss-side figures change quarterly and are compared against the public releases as they appear.

• Daily: attempts, right-party contacts, promises made, promises due today and kept, complaints and compliance exceptions. In the first weeks of a new team or script, this is the review that catches a problem before it reaches the roll rate.
• Weekly: RPC and PTP rates by bucket and by agent, promise-kept rate on last week's promises, dollars collected per collector hour, and note completeness on a sample.
• Monthly: delinquency by bucket on both bases, roll and cure rates by bucket, charge-off and net loss, recovery on prior vintages, and complaint rate by channel.
• Quarterly: the public comparisons. The New York Fed's Household Debt and Credit report for Q2 2026 was released on August 11, 2026; Experian's quarterly market data followed on August 27 via Auto Remarketing; the Federal Reserve Board's charge-off release was last updated on August 25, 2026; the CFPB's FDCPA annual report covering 2024 was published in November 2025. Each quarter, place your balance-weighted 90+ share beside the New York Fed's and your account-level 30- and 60-day rates beside Experian's, on their bases, not yours.
• For every cadence: the report header states the basis, population and period of each figure, and a definition does not change mid-year without a restated history. A KPI series that has been redefined twice is two series, and neither has a trend.

Where TEKS fits

TEKS Solutions provides first-party collections support: dedicated bilingual agents who work inside the client's own systems and under the client's name, from an operations floor in the Monterrey metropolitan area scheduled against U.S. Central hours, with a U.S. office in Arlington, Texas. Every program is run to Regulation F operating standards regardless of how the engagement is classified; that classification is determined with the client's counsel. The KPIs above are reported from the client's system on the client's definitions, with the contact-side figures reviewed daily through launch. TEKS does not publish rate cards; pricing follows the solution design.

The automotive finance and BHPH dealership pages describe typical programs, the Why TEKS page sets out the engagement model, the coverage page the hours, and the glossary defines the terms used here. A discovery call starts with the definitions in this guide and your current numbers on them.