关于「挽留对象与透明续费:风险乘可挽回性,以及不靠疏忽的收入」,《订阅叠砖、扣款挽回与年付拐点》给出的定量结论是什么?
结论:挽留对象按增益挑,不按风险挑:示意参数下,按风险挑与随机挑几乎一样,按估计增益挑能把净值从 19,328 提高到 29,686。最高风险的人平均增益只有中等风险者的三分之一。订阅收入里可能有三成多来自没察觉的订阅者,这层冰会在换卡、提醒与监管面前化掉;规划要用发提醒后的流失率。取消路径必须与订阅路径同等便捷,对比度、点按高度与步数都可以用程序检验。
数据口径:书中的订阅数、客单、流失率、挽回率、折现率与价格均为示意参数,全部计算由书中程序生成,不对应任何真实客户或成交数据;本书是非法律、会计、税务或投资建议,不承诺任何收入水平,「稳定」只指收入可预期并写明前提与失效条件。 第一至三章说明,稳态收入等于每月新增乘以客单再除以月流失率,新增定斜率,流失率定天花板;订阅者异质时总体留存随月龄上升,移位贝塔几何模型用几个月的队列数据就能外推;前两次续费是筛子,单次干预回报是后段的 1.40 倍,三年总增益却在长尾,示意参数下是 18,598 对 4,760;订阅层只在刚需使用次数明显高于值回票价的门槛时成立。
对照数据表
下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 19328, 29686
| rule | mean risk | saved | net value |
|---|---|---|---|
| top risk | 0.443 | 151.8 | 19,328 |
| random | 0.165 | 145.4 | 18,177 |
| estimated lift | 0.234 | 209.4 | 29,686 |
| true lift (bound) | 0.241 | 227.8 | 32,997 |
| mean lift of the 200 riskiest | 0.0340 | ||
| mean lift of risk 0.15 to 0.35 | 0.1011 |
查看完整验算程序
# Chapter 6. p61: target by lift, not by risk (the retention-futility
# pattern). 20,000 subscribers; churn risk p ~ Beta(1.2, 6). A retention
# contact lowers churn by lift = 0.25 * p * (1 - p) ** 3 * 4 plus noise,
# so the very highest-risk people are hard to move. Budget: contact
# 2,000 people at 4 units each; a kept subscriber is worth 180 units.
# The lift model is estimated with error (noise sd 0.02). Illustrative.
import math
import random
rng = random.Random(20261120)
PEOPLE, BUDGET, COST, WORTH = 20000, 2000, 4.0, 180.0
crowd = []
for _ in range(PEOPLE):
p = rng.betavariate(1.2, 6.0)
lift = max(0.0, 0.25 * p * (1 - p) ** 3 * 4 + rng.gauss(0, 0.01))
est = lift + rng.gauss(0, 0.02)
crowd.append((p, lift, est))
def outcome(chosen):
saved = math.fsum(lift for _, lift, _ in chosen)
return saved, saved * WORTH - COST * len(chosen)
by_risk = sorted(crowd, key=lambda x: -x[0])[:BUDGET]
by_est = sorted(crowd, key=lambda x: -x[2])[:BUDGET]
by_true = sorted(crowd, key=lambda x: -x[1])[:BUDGET]
rand = rng.sample(crowd, BUDGET)
print("rule mean risk saved net value")
RULES = (("top risk", by_risk), ("random", rand),
("estimated lift", by_est), ("true lift (bound)", by_true))
for name, group in RULES:
saved, net = outcome(group)
risk = math.fsum(p for p, _, _ in group) / len(group)
print("%-18s %9.3f %5.1f %10s" % (name, risk, saved,
"{:,.0f}".format(net)))
top = sorted(crowd, key=lambda x: -x[0])[:200]
print("mean lift of the 200 riskiest %.4f" % (
math.fsum(l for _, l, _ in top) / 200))
mid = [x for x in crowd if 0.15 <= x[0] <= 0.35]
print("mean lift of risk 0.15 to 0.35 %.4f" % (
math.fsum(l for _, l, _ in mid) / len(mid)))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《月度经常性收入叠砖、扣款失败自动挽回与年付现金流锁定母典》第 6 章「挽留对象与透明续费:风险乘可挽回性,以及不靠疏忽的收入」(版本 1.0.0). 5D Library. DOI 待发布
DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-367 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026367,
author = {Fan, Yafu},
title = {Stacking Monthly Recurring Revenue, Involuntary-Churn Dunning Recovery and Upfront Annual Cash Lock-In},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}