Find the cohort.
Check the denominator.
Which origination vintages and customer segments need a closer look? Inspect mature 180-day outcomes, predicted probabilities and the same fixed review capacity in one reproducible analysis.
Evaluation cutoff: 2025-07-01. Event: 90+ days past due within 180 days; this is the research label used by the source report.
Compare like with like
| Vintage | Segment | Loans | Events | Event rate | Mean PD | Rate − PD | Previous vintage Δ | Reviewed / loans | Review precision | Event capture | Principal |
|---|---|---|---|---|---|---|---|---|---|---|---|
| all | all | 480 | 107 | 22.3% | 18.8% | +3.5 pp | — | 48 / 480 | 52.1% | 23.4% | $7,977,052.50 |
| all | retail | 352 | 63 | 17.9% | 15.5% | +2.4 pp | — | 15 / 352 | 46.7% | 11.1% | $5,912,410.53 |
| all | small_business | 128 | 44 | 34.4% | 27.9% | +6.5 pp | — | 33 / 128 | 54.5% | 40.9% | $2,064,641.97 |
| 2024-07 | all | 80 | 17 | 21.2% | 18.6% | +2.7 pp | — | 8 / 80 | 25.0% | 11.8% | $1,406,219.88 |
| 2024-08 | all | 80 | 17 | 21.2% | 17.7% | +3.5 pp | +0.0 pp | 6 / 80 | 50.0% | 17.6% | $1,233,213.31 |
| 2024-09 | all | 80 | 15 | 18.8% | 16.8% | +1.9 pp | -2.5 pp | 7 / 80 | 71.4% | 33.3% | $1,349,401.92 |
| 2024-10 | all | 80 | 24 | 30.0% | 22.0% | +8.0 pp | +11.2 pp | 12 / 80 | 50.0% | 25.0% | $1,371,901.61 |
| 2024-11 | all | 80 | 16 | 20.0% | 20.6% | -0.6 pp | -10.0 pp | 10 / 80 | 60.0% | 37.5% | $1,347,411.92 |
| 2024-12 | all | 80 | 18 | 22.5% | 17.0% | +5.5 pp | +2.5 pp | 5 / 80 | 60.0% | 16.7% | $1,268,903.86 |
| 2024-07 | retail | 52 | 8 | 15.4% | 15.4% | +0.0 pp | — | 2 / 52 | 0.0% | 0.0% | $942,822.08 |
| 2024-07 | small_business | 28 | 9 | 32.1% | 24.5% | +7.7 pp | — | 6 / 28 | 33.3% | 22.2% | $463,397.80 |
| 2024-08 | retail | 60 | 10 | 16.7% | 14.2% | +2.5 pp | +1.3 pp | 1 / 60 | 0.0% | 0.0% | $909,200.76 |
| 2024-08 | small_business | 20 | 7 | 35.0% | 28.3% | +6.7 pp | +2.9 pp | 5 / 20 | 60.0% | 42.9% | $324,012.55 |
| 2024-09 | retail | 62 | 8 | 12.9% | 14.3% | -1.4 pp | -3.8 pp | 3 / 62 | 66.7% | 25.0% | $1,090,731.29 |
| 2024-09 | small_business | 18 | 7 | 38.9% | 25.4% | +13.5 pp | +3.9 pp | 4 / 18 | 75.0% | 42.9% | $258,670.63 |
| 2024-10 | retail | 60 | 15 | 25.0% | 18.4% | +6.6 pp | +12.1 pp | 4 / 60 | 50.0% | 13.3% | $1,034,713.40 |
| 2024-10 | small_business | 20 | 9 | 45.0% | 32.9% | +12.1 pp | +6.1 pp | 8 / 20 | 50.0% | 44.4% | $337,188.21 |
| 2024-11 | retail | 57 | 10 | 17.5% | 16.7% | +0.9 pp | -7.5 pp | 4 / 57 | 50.0% | 20.0% | $967,912.42 |
| 2024-11 | small_business | 23 | 6 | 26.1% | 30.4% | -4.3 pp | -18.9 pp | 6 / 23 | 66.7% | 66.7% | $379,499.50 |
| 2024-12 | retail | 61 | 12 | 19.7% | 14.0% | +5.6 pp | +2.1 pp | 1 / 61 | 100.0% | 8.3% | $967,030.58 |
| 2024-12 | small_business | 19 | 6 | 31.6% | 26.5% | +5.1 pp | +5.5 pp | 4 / 19 | 50.0% | 33.3% | $301,873.28 |
“Previous vintage Δ” compares the previous observed origination month in the same segment. A gap in months is not filled in. All outcomes use the same mature horizon. Blank denominators produce “—”, rather than a misleading 0%.
A fixed review queue
The source evaluation ranks calibrated probabilities, breaks ties by application ID and selects the preset capacity. Its historical review group has 52.1% precision. The 48-application queue contains scores and application details; future outcome labels are absent. This retrospective example does not establish the effect of an intervention.
Follow the calculations
Table schema · Cohort SQL · Queue SQL · Computed summary · Embedded source report · File digests
The SQL outputs are checked against separate Python counts and sums. The embedded source report makes the downloaded folder portable: creditvintage verify-analysis DIRECTORY reconstructs the analysis and this page from its verified source evidence.
Synthetic examples demonstrate data controls and calculation methods. Segment differences and month-to-month changes are descriptive; they do not establish causes, significance or real borrower risk. Loss figures in the CSV use principal × probability/event × the recorded LGD assumption, not realized recoveries or accounting provisions. Keep private borrower exports private.