Market data
Coverage, provenance, quality gates, and model benchmarks.
This page reports coverage, provenance, and automated quality checks from the same artifact that powers the public price check.
Pipeline health
All release gates passed
- Schema contract PASS
- Duplicate composite keys0
- Percentile-order violations0
- Rejected market cells0
- Minimum cell sample n≥ 10
Artifact provenance
Reproducible from the released source
- Retrieved
- Aug 29, 2026
- Licence
- CC BY-NC 4.0
- Price basis
- Dealer asking prices, not transactions
- SHA-256 (abbrev.)
- 93f41bf1…05702c8e
Consumer adjustment model
Trained on completed auction outcomes.
Canadian anchor. Learned auction residual. Read the model card summary →
Research benchmark
Evaluated against a declared baseline.
Five-fold WAPE (grouped) · Five folds. Zero make-model overlap.
Segment validation
Where the error concentrates.
| Segment | n | Baseline MAE | Model MAE | Baseline WAPEBase WAPE | Model WAPE | Model 80% coverageCoverage |
|---|---|---|---|---|---|---|
| Auction grade | ||||||
| Salvage | 69 | $3,724.57 | $2,249.09 | 156.18% | 94.31% | 15.94% |
| Extra Rough | 1,142 | $1,796.93 | $1,317.73 | 34.99% | 25.66% | 49.56% |
| Rough | 9,184 | $1,255.62 | $1,095.69 | 18.42% | 16.07% | 69.54% |
| Average | 18,546 | $1,259.39 | $1,145.42 | 13.25% | 12.05% | 81.35% |
| Clean | 9,607 | $1,138.54 | $1,351.31 | 7.37% | 8.75% | 90.75% |
| Extra Clean | 584 | $1,460.93 | $1,609.11 | 7.93% | 8.74% | 91.44% |
| Sale year | ||||||
| 2009 | 36,116 | $1,282.86 | $1,213.41 | 12.80% | 12.11% | 78.85% |
| 2010 | 3,016 | $880.85 | $1,015.96 | 6.30% | 7.27% | 93.73% |
| Price band | ||||||
| $0-5k | 8,496 | $991.42 | $722.48 | 30.28% | 22.07% | 58.07% |
| $5-10k | 12,544 | $1,193.87 | $983.95 | 16.14% | 13.30% | 80.55% |
| $10-20k | 14,737 | $1,342.76 | $1,442.71 | 9.65% | 10.37% | 89.23% |
| $20k+ | 3,355 | $1,729.14 | $2,129.83 | 7.40% | 9.11% | 92.88% |
| Odometer delta vs peers | ||||||
| S1 | 6,522 | $1,745.05 | $1,581.84 | 14.35% | 13.00% | 74.10% |
| S2 | 6,522 | $1,246.29 | $1,147.52 | 11.78% | 10.84% | 79.58% |
| S3 | 7,007 | $991.86 | $984.87 | 9.76% | 9.70% | 82.49% |
| S4 | 6,037 | $876.59 | $944.66 | 9.04% | 9.74% | 84.81% |
| S5 | 6,522 | $977.19 | $1,053.31 | 10.23% | 11.03% | 82.83% |
| S6 | 6,522 | $1,665.71 | $1,473.95 | 17.02% | 15.06% | 76.34% |
| Peer count | ||||||
| T1 | 13,070 | $1,401.74 | $1,203.91 | 15.68% | 13.46% | 74.22% |
| T2 | 13,232 | $1,258.61 | $1,161.13 | 12.99% | 11.98% | 78.24% |
| T3 | 12,830 | $1,092.26 | $1,230.58 | 8.81% | 9.93% | 87.69% |
MAE/WAPE: lower is better — ▾ model beats baseline, ▴ model trails; coverage red when below 80%.
80% coverage is the share of each segment's actual sale prices inside the global temporal-test interval [P10, P90] of log(actual / model prediction); its width is 45.86% of the central estimate, so it does not re-fit per segment. The model improves WAPE in 13 of 21 segments; the remaining rows are where the peer baseline already prices well.
Coverage is not uniform. Cells below 10 vehicles are suppressed.
ABBCMBNBNLNSONPEQCSK