B2B 分销 · 产品营销
每年分类 21,000 个产品，首次建议准确率 82%。
Manutan 需将数千个新品归入内部品类。带三级评分与 HITL 的引擎已在生产运行。
关键指标
- 客户
- Manutan
- 行业
- B2B 分销
- 职能
- 产品营销
- 状态
- 已投产
- 部署
- 客户平台
背景
Every year, thousands of new products from many suppliers must be filed into article sub-families — the taxonomy that structures offer, merchandising and search.
问题
- Manual classification, heterogeneous and sometimes multilingual descriptions.
- Errors hitting findability and sales. No confidence score on manual assignments.
方案
Multi-supplier ingestion, semantic analysis, embedding match, high / medium / low confidence scoring, HITL loop, capitalization on every correction.
调用的 NEXA 组件
| 组件 | 作用 |
|---|---|
| Arbitration algorithm | Multi-criteria semantic assignment |
| Evidence Panel | Score, alternatives, justification |
| HITL | Merchandising validation / correction |
| AI Knowledge Vault | Validated patterns for the next wave |
| Cockpit | Green / amber / red volumes |
影响
| 维度 | 具体影响 |
|---|---|
| Precision | 82% correctly classified on the first proposal (high confidence). |
| Volume | ~21,000 products/year from 39 supplier files. |
| Adoption | Merchandising team equipped across the full flow. |
« 每年手工分类 21,000 个品项已不可持续。第一次建议带着分数到来：我们验证，不再事后拼凑。 »
82%
首次建议准确率