PHILLIPS INSIGHTS / EXECUTIVE SERVICE PREVIEW
Where is risk building?
What should management do next?
A worked auto finance diagnostic that connects portfolio deterioration, origination quality, collections flow, and economics to an executive decision agenda.
Open the executive diagnosticInteractive synthetic case · five analytical views · transparent calculations
ILLUSTRATIVE CREDIT COMMITTEE BRIEF / Q3 2026
Risk is building across the portfolio.
The headline rate is only the start.
A fictional $500 million auto portfolio, designed to demonstrate the depth of a Phillips Insights diagnostic. All portfolio figures, cohorts, scenarios, and action thresholds below are synthetic. Public-source observations appear separately further down the page.
THE EXECUTIVE READOUT
Three signals. Three decisions to investigate.
The demonstration points to a combined mix and performance problem. Neither growth restraint nor tighter credit policy should be prescribed from a portfolio average alone.
Mix explains 47% of the delinquency increase.
60+ DPD rises from 2.57% to 3.79%. Holding prior tier rates fixed attributes 56.9 bps to changing balance mix; the remaining 65.3 bps comes from higher rates within tiers.
Decision implication: Separate growth allocation from within-segment deterioration before setting a response. This arithmetic decomposition is not causal proof of underwriting changes.
Newer vintages deteriorate at the same age.
The 2025 Q3 cohort reaches 2.80% cumulative net loss at 12 months on book, versus 1.50% for 2024 Q3 at the same age: a 130-basis-point gap. Later outcomes for the newer cohort are unobserved.
Decision implication: Review matched-age performance by channel, term, and LTV. The cohorts also differ in composition, so the gap does not isolate a credit-policy effect.
Margin headroom is thin relative to loss sensitivity.
The contribution proxy is 2.38%, only 38.0 bps above a hypothetical 2.00% hurdle. A 25% relative NCO uplift plus 50 basis points of funding pressure reduces it to 1.04%.
Decision implication: Evaluate the economics of growth and funding together. The scenario is a mechanical sensitivity—not a calibrated forecast, CECL estimate, or capital model.
Proposed management agenda for this fictional portfolio
Credit / riskReconcile cohort and segment definitions; test whether deterioration persists within comparable origination cells.
Finance / treasuryValidate yield and funding attribution; compare growth plans against loss and funding sensitivities.
Servicing / collectionsTrack early delinquency exits and forward rolls; test treatments against a control group before claiming incremental benefit.
Inspect the metric dictionary, assumptions, and demonstration data
- Portfolio scope
- Fictional US auto-loan book; Q3 2026 current snapshot versus a prior snapshot. $500m average and ending principal assumed equal for annualized economics. No named lender or client is represented.
- Delinquency bridge
- Prior DPD = Σ prior weight × prior rate. Mix = Σ (current weight − prior weight) × prior rate. Within-tier = Σ current weight × (current rate − prior rate). Current minus prior equals mix plus within-tier, before rounding. All four tiers are mutually exclusive and exhaustive within the example.
- Loss and economics basis
- NCO = quarterly net charge-offs ÷ average principal × 4. Portfolio NCO and yield are current-weighted tier rates. Annualized dollars use $500m average principal; they are run-rate conversions, not one-year forecasts. Tier yields are synthetic realized yields, not APR offers.
- Independent analytical exhibits
- The vintage and migration datasets demonstrate different analytical lenses. They are not reconciled to each other or to a single underlying loan tape. No causal model, prediction accuracy, client benefit, or optimized policy is claimed.
- Validation for production use
- Reconcile to servicing and finance totals; freeze observation dates; inspect missingness, definitions, prepayments, recoveries, and selection effects; validate cohort comparability and model performance before acting.
PUBLIC DISCLOSURES / A SEPARATE LENS
Real observations.
Careful comparisons.
A research view starts with dated sources and precise definitions. These public facts are separate from the fictional portfolio above.
US auto-loan balances
The New York Fed reported a $28 billion increase in auto-loan balances during Q2 2026. Its summary also noted an uptick in transitions into early auto delinquency.
Analytical use: Broad credit-cycle and market context. National credit-panel measures do not establish a like-for-like specialty-lender peer group.
Source: Q2 2026 Household Debt and Credit report summary ↗1.57%Retail auto annualized NCO
4.80%Retail auto 30+ DPD, including nonaccrual
Ally reported NCO down 18 basis points year over year and 30+ delinquency down 8 basis points. These are disclosed portfolio results, not the demonstration’s assumptions.
Comparability limit: 30+ DPD is not 60+ DPD; annualized NCO uses average receivables. Credit mix, seasoning, accounting policies, and portfolio scope must be aligned before a peer comparison.
Source: Ally Q2 2026 release, pages 3 and 5 ↗Historical public-source snapshot: quarter ended June 30, 2026; sources checked October 17, 2026. No affiliation or endorsement is implied. Reported observations are not used to calibrate the synthetic case or establish a lender ranking.
THE ENGAGEMENT / FROM SIGNAL TO EVIDENCE
What you would take
into the next credit committee.
A trusted analytical base.
Servicing-to-finance reconciliation, explicit metric definitions, cohort eligibility rules, and a documented observation window. Exceptions are visible before conclusions are drawn.
Deliverable: reconciled scorecard, data dictionary, exception log.A defensible driver analysis.
Mix-versus-rate bridges, matched-age vintage views, channel and collateral segmentation, and migration diagnostics. Separate supported findings from hypotheses that need testing.
Deliverable: analytical exhibits, driver assessment, limitations register.A decision-ready set of options.
Contribution sensitivities, operating tradeoffs, proposed measurement plans, and decision ownership. If predictive modeling is appropriate, agree the target, validation design, and monitoring criteria first.
Deliverable: scenario workbook, test plan, executive decision memo.A repeatable management cadence.
Hand over metric logic, source lineage, refresh controls, and reporting ownership so the analysis can be repeated and challenged as the portfolio changes.
Deliverable: refresh specification, monitoring pack, prioritized roadmap.Designed for the question you need to answer.
Support across the auto lending spectrum, with particular focus on smaller midprime and subprime lenders. A diagnostic’s scope depends on data readiness, portfolio complexity, and the management decision. Private loan data would be exchanged through an agreed secure process—not the public inquiry form.
Discuss a portfolio diagnosticANALYTICAL DISCIPLINE
Evidence, assumptions, and decisions stay traceable.
This preview demonstrates analytical depth, not a validated production credit model. A client engagement would address data lineage, independent challenge, performance validation, and ongoing monitoring appropriate to the intended use. Any changes affecting underwriting, pricing, or collections would require the lender’s governance and applicable compliance review.
Methodological reference: Federal Reserve SR 26-2, revised guidance on model risk management (April 17, 2026) ↗. This reference does not imply certification, supervisory approval, or applicability to every lender.
START WITH THE DECISION
Bring the portfolio question
your current reporting can’t resolve.
We’ll discuss the evidence, data, and analytical work needed to get to an answer.
Discuss your portfolio