关于「扣款失败:被动流失的原因分类与重试排程」,《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:第四章把被动流失按余额不足、卡片过期、风控拒绝与硬拒绝分类,在卡组织公开上限之内排重试,全部失败的挽回率从 0.232 升到 0.627,挽回七成可把天花板抬到 1.762 倍;硬拒绝不重试,卡片信息只由持牌支付机构保管。

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。

English version

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: 0.232, 0.627

reasonnaive 0,1,2,3best days (cap 4)bestcap 2
insufficient0.2807,9,10,130.8900.751
expired0.21810,11,12,130.6600.659
risk0.2861,2,3,40.3440.190
hard (0.12)never retried: ask the holder for a new card
all failures, naive schedule0.232
all failures, best schedules0.627
查看完整验算程序
# Chapter 4. p42: best retry days by decline reason under an attempt cap.
# Day 0 is the failed renewal charge; retries may use days 0 to 13 and
# at most 4 attempts (an internal cap set well inside the public network
# limits, which change over time). Hard declines are never retried.
# Insufficient funds: funds arrive on day D (illustrative weights; 0.08
# of cases fall after day 13); before D an attempt succeeds 0.03, from D
# on 0.85 * 0.93 ** (t - D); weekend days 4, 5, 11, 12 multiply by 0.9.
# Expired card: the holder updates details after a notice with daily
# hazard 0.08; an attempt after the update succeeds 0.95, before it 0.
# Risk decline: 0.10 per attempt, 0.02 if made on day 0. Because only a
# failure leads to the next attempt, a fixed day set is a full policy,
# so exhaustive enumeration with itertools is exact. All illustrative.
# Chapter 4 hypothesis: funds-day timing lifts recovery by >= 0.20.
import itertools
import math

DAYS, CAP = 14, 4
W = (0.02, 0.05, 0.06, 0.07, 0.08, 0.10, 0.12, 0.12, 0.10, 0.08, 0.05,
     0.03, 0.02, 0.02)
LATE = 1.0 - math.fsum(W)
WEEKEND = (4, 5, 11, 12)


def p_if(t, d):
    p = 0.03 if t < d else 0.85 * 0.93 ** (t - d)
    return p * (0.9 if t in WEEKEND else 1.0)


def rec_if(days):
    terms = []
    for d, w in enumerate(W):
        miss = 1.0
        for t in days:
            miss *= 1.0 - p_if(t, d)
        terms.append(w * (1.0 - miss))
    miss = 1.0
    for t in days:
        miss *= 0.97
    terms.append(LATE * (1.0 - miss))
    return math.fsum(terms)


def rec_exp(days):
    terms = []
    for u in range(DAYS):
        w = 0.08 * 0.92 ** u
        miss = 1.0
        for t in days:
            if t > u:
                miss *= 0.05
        terms.append(w * (1.0 - miss))
    return math.fsum(terms)


def rec_risk(days):
    miss = 1.0
    for t in days:
        miss *= 1.0 - (0.02 if t == 0 else 0.10)
    return 1.0 - miss


def best(fn, cap):
    top = ((), 0.0)
    for k in range(1, cap + 1):
        for days in itertools.combinations(range(DAYS), k):
            v = fn(days)
            if v > top[1] + 1e-12:
                top = (days, v)
    return top


MIX = (("insufficient", 0.45, rec_if), ("expired", 0.25, rec_exp),
       ("risk", 0.18, rec_risk))
naive = (0, 1, 2, 3)
print("reason         naive 0,1,2,3   best days (cap 4)   best   cap 2")
tot_n, tot_b = [], []
for name, share, fn in MIX:
    days, v = best(fn, CAP)
    v2 = best(fn, 2)[1]
    tot_n.append(share * fn(naive))
    tot_b.append(share * v)
    print("%-13s  %13.3f   %-18s  %5.3f  %5.3f" % (
        name, fn(naive), ",".join(str(d) for d in days), v, v2))
print("hard (0.12)    never retried: ask the holder for a new card")
print("all failures, naive schedule   %.3f" % math.fsum(tot_n))
print("all failures, best schedules   %.3f" % math.fsum(tot_b))

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 4 章「扣款失败:被动流失的原因分类与重试排程」(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}
}

本书其他章节结论