Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →银行经营指标变动归因拆解。当用户要求分析存款下降原因、贷款增长驱动因素、利润变动归因、净息差变动拆解、中间业务收入波动归因、不良率变动成因时使用。按机构/产品/客群/期限等维度多维归因。
.claude/skills/aifinlab-business-metrics-attribution/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 47% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 27% | 0% |
银行经营指标发生变化时（如利润同比下降、存款环比增长放缓、中间业务收入波动等），管理层需要快速定位驱动因素。传统分析依赖人工逐项拆解，效率低且容易遗漏交叉影响。本技能将指标变动按维度（机构、产品、客群、期限等）进行结构化归因拆解，输出"结论先行、数据支撑"的归因报告。
| 输入项 | 说明 | 是否必须 | |--------|------|----------| | 指标名称 | 需要归因的经营指标（如"利润总额"、"存款余额"） | 必须 | | 当期数据 | 当期各维度的明细数据 | 必须 | | 基期数据 | 对比基期的同口径数据 | 必须 | | 归因维度 | 按哪些维度拆解（机构/产品/客群/期限等） | 可选，默认按机构+产品 | | 分析口径 | 同比/环比/较年初 | 可选，默认同比 |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 56,940 | 48,391 | -15% | 1 | 1 | 0% | 4,600 | 3,882 | -16% | 0 | 0 | — |
case-02 | fail→fail | 56,396 | 28,014 | -50% | 1 | 1 | 0% | 4,258 | 5,692 | +34% | 0 | 0 | — |
case-03 | fail→pass | 18,672 | 19,770 | +6% | 1 | 1 | 0% | 3,584 | 5,242 | +46% | 0 | 0 | — |
case-04 | pass→pass | 17,410 | 15,339 | -12% | 1 | 1 | 0% | 2,789 | 3,877 | +39% | 0 | 0 | — |
case-05 | pass→pass | 16,171 | 17,308 | +7% | 1 | 1 | 0% | 2,771 | 4,270 | +54% | 0 | 0 | — |
case-06 | pass→pass | 15,736 | 18,757 | +19% | 1 | 1 | 0% | 2,647 | 4,318 | +63% | 0 | 0 | — |
case-07 | pass→pass | 19,289 | 15,738 | -18% | 1 | 1 | 0% | 3,377 | 4,020 | +19% | 0 | 0 | — |
case-08 | pass→pass | 26,779 | 26,544 | -1% | 1 | 1 | 0% | 5,098 | 6,059 | +19% | 0 | 0 | — |
case-09 | pass→pass | 20,492 | 41,933 | +105% | 1 | 1 | 0% | 2,688 | 3,186 | +19% | 0 | 0 | — |
case-10 | pass→pass | 19,799 | 18,444 | -7% | 1 | 1 | 0% | 3,324 | 4,393 | +32% | 0 | 0 | — |
case-11 | pass→pass | 21,860 | 17,435 | -20% | 1 | 1 | 0% | 3,299 | 4,010 | +22% | 0 | 0 | — |
case-12 | pass→pass | 18,946 | 14,621 | -23% | 1 | 1 | 0% | 3,089 | 3,608 | +17% | 0 | 0 | — |
case-18 | pass→pass | 19,586 | 15,567 | -21% | 1 | 1 | 0% | 3,383 | 4,076 | +20% | 0 | 0 | — |
case-13 | fail→pass | 13,520 | 13,205 | -2% | 1 | 1 | 0% | 2,130 | 3,138 | +47% | 0 | 0 | — |
case-14 | pass→pass | 20,860 | 18,151 | -13% | 1 | 1 | 0% | 3,494 | 4,083 | +17% | 0 | 0 | — |
case-15 | fail→pass | 20,595 | 18,723 | -9% | 1 | 1 | 0% | 2,967 | 3,482 | +17% | 0 | 0 | — |
case-16 | pass→pass | 12,169 | 18,001 | +48% | 1 | 1 | 0% | 2,227 | 4,177 | +88% | 0 | 0 | — |
case-17 | pass→pass | 12,492 | 16,776 | +34% | 1 | 1 | 0% | 2,173 | 3,938 | +81% | 0 | 0 | — |
case-19 | pass→pass | 19,671 | 17,775 | -10% | 1 | 1 | 0% | 3,191 | 4,427 | +39% | 0 | 0 | — |
case-20 | pass→pass | 19,401 | 18,840 | -3% | 1 | 1 | 0% | 3,113 | 4,356 | +40% | 0 | 0 | — |
case-21 | fail→pass | 10,362 | 7,088 | -32% | 1 | 1 | 0% | 1,626 | 2,057 | +27% | 0 | 0 | — |
case-22 | fail→pass | 42,643 | 17,392 | -59% | 1 | 1 | 0% | 3,812 | 3,645 | -4% | 0 | 0 | — |
case-23 | pass→pass | 11,354 | 15,518 | +37% | 1 | 1 | 0% | 1,537 | 3,294 | +114% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +26 percentage points is the difference between those two pass rates over the 23 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.