捐赠人留存分析代理

概览

价值

功能说明

• 收入线背后的留存

  收入保持平稳而支持者基础在崩塌，因为新捐赠者补上了流失的，总数不动。

• 找出损失发生在哪里

  几乎总是在首次捐赠之后、在第一年里。那与中途流失需要的是不同的修正。

• 读流失之前的经历

  有没有发出感谢、有没有人说过这笔钱做了什么，以及下一次募捐是否先于这两者到来。

• 把扣款失败与主动决定分开

  一张过期的卡不等于一位流失的支持者。没有人作过任何决定，也没有人去跟进。

• 衡量第二次捐赠的价值

  获取成本是首次捐赠的好几倍。全部经济逻辑都在它之后的那些捐赠里。

• 不给任何个人评分

  它不建立财富或捐赠能力评级。按「可能被劝说捐多少」给支持者排名，是另一种带着自身伦理问题的产品。

这会改变什么

运行方式

该智能体通过 PROMIVO 运行时在服务器端运行。每次运行都会逐步记录，每次工具调用在执行前都会进行权限检查。

1. 1输入按智能体的输入结构进行校验
2. 2模型claude-sonnet-5，并自动回退
3. 3工具extract_structured_data
4. 4权限自动执行 READ 级操作；更高的等待你批准
5. 5输出符合输出结构的结构化结果
6. 6日志记录 token、耗时和每一个步骤

演示

演示数据Illustrative sample output, abridged.

{
  "period": "2022–2026",
  "currency": "USD",
  "supporters": [],
  "suppressionThreshold": 10
}

{
  "bySource": [
    "One acquisition source costs the least per supporter and retains at 11% to year two. It is the most expensive channel you have once the second gift is counted."
  ],
  "escalate": true,
  "whatWorks": [
    "Supporters who received a handwritten acknowledgement on a first gift above a threshold retained at 58%. Small volume, and worth knowing before it is discontinued."
  ],
  "disclaimer": "An analysis of your own supporter records. No wealth, capacity or propensity score has been built for anyone; no supporter has been contacted or identified; groups below ten are suppressed; modelled uplift is derived from your own history rather than forecast; and this is not advice on fundraising regulation.",
  "modelledUplift": [
    "Bringing first-to-second gift conversion from 19% to the 2023 level of 26% is worth roughly 210,000 a year at your own average gift values. Modelled from your history, not a forecast."
  ],
  "cohortRetention": [
    {
      "cohort": "2023",
      "supporters": 4120,
      "retainedYear1": 34,
      "retainedYear2": 21,
      "firstToSecondGiftPercent": 26
    },
    {
      "cohort": "2025",
      "supporters": 5010,
      "retainedYear1": 27,
      "retainedYear2": 0,
      "firstToSecondGiftPercent": 19
    }
  ],
  "escalationReason": "First-to-second gift conversion is 19%, below threshold and falling across cohorts, and 23% of regular gift lapses are mechanical with no follow-up recorded.",
  "groupsSuppressed": 5,
  "incomeVersusBase": "Income is within 3% of four years ago. The supporter base is down 41% over the same period. New donors have been replacing lapsed ones and the income line shows none of it.",
  "mechanicalLapses": "23% of regular gift lapses are expired cards or failed instructions. Nobody in that group decided to stop supporting you, and no record shows any of them being contacted.",
  "stewardshipFindings": [
    "Of supporters who lapsed after one gift, 71% received no thank-you within 14 days.",
    "44% received a second ask before any communication about what their first gift did.",
    "Supporters who received an impact update within 60 days gave again at more than twice the rate."
  ]
}

使用条件

使用场景

价格

常见问题

评价

客户运行过该智能体后，评价功能将开放。