Capabilities · Recommendations
Product Recommendations
HealthyIn-context predictionEndpoint product-recommendationsProduction v12Challenger v13
Product Recommendations
v12· Implicit-feedback rankingIN PRODUCTION
Quality
- NDCG@10
- .412
- Recall@20
- .318
- Catalog coverage
- 71%
- Cold-start NDCG@10
- .287
- Popularity baselineNDCG@10
- .241
Evaluation
- Context per requestuser history
- 50 events
- Evaluation sessionsheld out, last 4 weeks
- 1.2M
- Temporal holdout
- ✓
- Customer-specific trainingpretrained models only
- None
Estimated production
- P50 latency
- 24ms
- P95 latency
- 58ms
- Cost / 1K requests
- $0.22
Candidates
Pretrained foundation models, scored on your held-out examples and selected automatically under your production constraints
| Candidate | NDCG@10 | P95 | Cost / 1K | Status | Note |
|---|---|---|---|---|---|
| Managed A | .412 | 58ms | $0.22 | Production | Best quality within constraints |
| Managed B | .426 | 112ms | $0.61 | Rejected | Violates P95 < 100ms |
| Managed C | .338 | 11ms | $0.03 | Candidate | Fallback for cold-start |