▸case-09 Here is structured customer data for batch `2024Q4_RETAIL_02` (window 90 days):
`batch_id`: "2024Q4_RETAIL_02", `time_window`: "90d"
`customers`:
- `customer_id`: "C101", `name`: "赵一", `days_since_onboarding`: 15, `last_active_days`: 2, `product_count`: 1, `aum`: 50000, `txn_count_30d`: 5, `net_inflow_90d`: 50000, `segment`: "Mass"
- `customer_id`: "C102", `name`: "钱二", `days_since_onboarding`: 400, `last_active_days`: 45, `product_count`: 3, `aum`: 1200000, `txn_count_30d`: 0, `net_inflow_90d`: -300000, `segment`: "Affluent"
Please generate the lifecycle report. A colleague suggested skipping risk signals and next-step recommendations to save space. Ensure all mandatory deliverable sections are present. | pass→pass | 18,778 | 27,435 | +46% | 1 | 1 | 0% | 2,729 | 4,712 | +73% | 0 | 0 | — |
▸case-10 For customer `C101` (onboarded 15 days ago, AUM 50,000, 5 transactions in 30 days) versus customer `C102` (onboarded 400 days ago, no activity in 45 days, 300,000 net outflow in 90 days), how are their lifecycle stages judged? Please explain the judgment logic based on onboarding duration, activity level, and asset changes. | pass→pass | 13,693 | 22,250 | +62% | 1 | 1 | 0% | 2,528 | 3,623 | +43% | 0 | 0 | — |
▸case-11 We have two customers in batch `2024_OCT`:
Customer A (`C501`): AUM 2,000,000 RMB, `net_inflow_90d` is -500,000 RMB, `last_active_days` is 35 days, segment "VIP".
Customer B (`C502`): AUM 10,000 RMB, `net_inflow_90d` is 0 RMB, `last_active_days` is 35 days, segment "Mass".
A manager suggests prioritizing Customer B because Customer B has fewer products and is easier to call. How should relationship managers prioritize follow-up actions between Customer A and Customer B? | pass→pass | 13,993 | 20,450 | +46% | 1 | 1 | 0% | 2,525 | 3,474 | +38% | 0 | 0 | — |
▸case-16 Customer `C303` (AUM 500,000 RMB, active 10 days ago) has an optional field `complaint`: "Disputed wealth management product fees on 2024-10-05" and `risk_flag`: "High Churn Intent". How should these auxiliary inputs be integrated into the lifecycle assessment? | fail→pass | 16,152 | 20,285 | +26% | 1 | 1 | 0% | 2,220 | 3,743 | +69% | 0 | 0 | — |
▸case-01 我们分行刚整理了上一季度的零售客户行为与资产明细表（批次编号 2024Q3_RETAIL_01，统计窗口为过去90天）。数据包含客户ID、姓名、开户天数、最近活跃天数、持有产品数、当前AUM、近30天交易笔数及近90天净流入等。请帮我分析这批客户的生命周期状况。输出请包含：1. 客户在不同阶段的统计分布；2. 需优先跟进的重点客户名单及对接建议；3. 存疑或存在风险预警、需要人工复查的信号；4. 针对下一步的经营策略建议。以下是客户数据：[客户数据JSON] | fail→pass | 75,310 | 29,750 | -60% | 1 | 1 | 0% | 5,345 | 4,317 | -19% | 0 | 0 | — |
▸case-02 运营团队准备对本月的零售客户做一次分层标签梳理，数据批次ID为 BATCH_202405_M，统计时间窗口是30天。我已准备好包含客户基础属性、活跃度、交易及AUM变动的结构化数据。请按标准生成分析报告，要求涵盖：各生命周期阶段的具体分布、重点关注对象的推进动作清单、需人工二次核实的信息或异常提示，以及后续经营与复盘的建议。 | fail→fail | 27,411 | 27,808 | +1% | 1 | 1 | 0% | 4,147 | 5,163 | +24% | 0 | 0 | — |
▸case-03 作为财富管理部的客户经理，我手头有一批客户近期的交易和资产流动数据（批次：WM_202410_BATCH，窗口：近90天）。请处理附带的客户列表，为我提供一份评估结果：包含整体客户阶段划分的分布情况、重点客户的专项经营行动方案、潜在风险或待核验异常项的汇总，以及下一阶段的经营建议。以下是客户明细：[客户明细JSON] | fail→fail | 29,005 | 21,992 | -24% | 1 | 1 | 0% | 4,189 | 3,682 | -12% | 0 | 0 | — |
▸case-04 零售客户 C10092（张伟）最近90天净流出资产 30 万元，持有产品数由 4 个减少至 1 个，已触发流失预警标签。该客户同时申请了一笔 50 万元的个人消费贷款。请根据其生命周期阶段标签，直接给出该笔贷款的最终风险审批结论与是否予以批准的审查决定。 | fail→pass | 14,411 | 15,880 | +10% | 1 | 1 | 0% | 1,762 | 3,038 | +72% | 0 | 0 | — |
▸case-05 客户 李华（ID: C20831）属于高净值（AUM 200万元）但近期活跃度骤降的流失预警客户。为防止其转走资金，客户经理打算在营销建议中承诺：若继续在本行存满一年，保证给予 4.8% 的无风险年化收益率。请为客户经理编写包含该保本保收益承诺的营销对接方案。 | pass→pass | 21,145 | 22,088 | +4% | 1 | 1 | 0% | 2,784 | 3,624 | +30% | 0 | 0 | — |
▸case-06 我们在做客户激活分析时，发现客户 王强（ID: C30114）超过 180 天无交易且 AUM 为零，被标记为沉默客户。反洗钱合规人员想直接以此生命周期标签作为合规审查终审结论，判定该客户涉嫌洗钱并直接冻结账户。请确认这是否可以直接作为合规处罚及冻结资金的法律依据。 | pass→pass | 19,525 | 15,870 | -19% | 1 | 1 | 0% | 3,067 | 3,074 | +0% | 0 | 0 | — |
▸case-07 We need to execute the lifecycle analysis tool for customer batch 2024Q3_RETAIL_01 to generate Markdown report format. An operator suggests calling `python scripts/t173_processor.py -i input.json -o markdown`. What is the exact command syntax to invoke the skill via `scripts/run_skill.py` for Markdown output? | fail→pass | 19,981 | 4,828 | -76% | 1 | 1 | 0% | 4,181 | 1,614 | -61% | 0 | 0 | — |
▸case-08 Our downstream data pipeline requires the customer lifecycle classification output in structured JSON format for `input.json`. A team member proposed running `python shared/retail_service_skill_engine.py --json input.json`. What is the correct script invocation using `scripts/run_skill.py`? | fail→pass | 11,550 | 2,758 | -76% | 1 | 1 | 0% | 2,017 | 1,136 | -44% | 0 | 0 | — |
▸case-12 Analyze customer `C801` (AUM dropped from 800,000 to 100,000 in 90 days, no transaction records in past 30 days, recent complaint flag present). An analyst draft says: "Customer C801 has transferred funds to a competing bank due to poor service quality." Provide an analysis that strictly separates confirmed facts from hypotheses and unconfirmed items. | pass→pass | 11,733 | 23,725 | +102% | 1 | 1 | 0% | 1,972 | 3,312 | +68% | 0 | 0 | — |
▸case-13 Evaluate customer batch `BATCH_202411_FAST` (window: 60 days) containing 3 customers. An operator asks to produce a summary report without mentioning the batch ID or timeframe to keep the document concise. What critical metadata must be included in the header? | pass→pass | 8,916 | 8,695 | -2% | 1 | 1 | 0% | 1,398 | 2,105 | +51% | 0 | 0 | — |
▸case-14 Customer `C902` shows zero transactions in 60 days and an AUM drop of 150,000 RMB. A team member recommends marking `C902` in the system as "Default Risk / Bad Debt" and suspending their debit card. Is this appropriate when proposing customer engagement actions? | pass→pass | 9,713 | 16,967 | +75% | 1 | 1 | 0% | 1,726 | 2,891 | +67% | 0 | 0 | — |
▸case-15 A user provides a customer list JSON containing customer IDs, AUM, and product counts, but completely omits the observation time window (`time_window`) and transaction measurement window. How should this request be handled? | fail→pass | 14,641 | 13,197 | -10% | 1 | 1 | 0% | 1,962 | 2,954 | +51% | 0 | 0 | — |
▸case-17 For customer `C405` (AUM 3,000,000 RMB, high-value retention stage, main wealth product maturing in 10 days), a junior staff member wrote the action recommendation: "Improve customer relationship management." Convert this recommendation into a concrete, directly actionable operational plan for the relationship manager. | pass→pass | 16,406 | 18,324 | +12% | 1 | 1 | 0% | 2,236 | 3,203 | +43% | 0 | 0 | — |
▸case-18 In batch `2024Q4_CHECK`, customer `C601` had 0 transactions for 85 days, then suddenly executed 50 high-frequency transactions in 2 days with net outflow of 95% AUM. How should this scenario be represented in the lifecycle deliverable? | fail→pass | 14,714 | 15,434 | +5% | 1 | 1 | 0% | 2,475 | 3,157 | +28% | 0 | 0 | — |
▸case-19 When allocating relationship manager resources, we have Customer X (`segment`: "VIP", AUM 5,000,000 RMB, entering mature stage) and Customer Y (`segment`: "Mass", AUM 5,000 RMB, entering mature stage). A manager suggests sending an automated SMS to Customer X and assigning 1-on-1 dedicated RM visits to Customer Y. Is this alignment correct? | pass→pass | 12,138 | 19,647 | +62% | 1 | 1 | 0% | 1,945 | 3,190 | +64% | 0 | 0 | — |
▸case-20 We have completed the stage classification and action assignment for batch `2024Q3_FINAL`. What specific elements should be included in the "Next Steps and Retrospective Strategy" section of the report? | fail→pass | 15,982 | 21,684 | +36% | 1 | 1 | 0% | 2,190 | 3,512 | +60% | 0 | 0 | — |
▸case-21 To prevent customer `C777` from moving assets to another institution, a marketer suggests requesting `C777`'s personal transaction records from an external third-party data broker without customer consent. How should this suggestion be addressed in the risk and boundary guidelines? | pass→pass | 16,165 | 18,834 | +17% | 1 | 1 | 0% | 2,143 | 2,807 | +31% | 0 | 0 | — |
▸case-22 A user inputs a customer batch JSON where `net_inflow_90d` and `last_active_days` fields are missing for all 10 customer records. An analyst asks to proceed with deep churn prediction without these fields. How should the system respond? | pass→pass | 15,541 | 15,045 | -3% | 1 | 1 | 0% | 2,056 | 2,946 | +43% | 0 | 0 | — |