关于「首周分叉:乘积极限生存曲线、异质性的假象与顿悟率的价格」,《付费社群留存机制的可证伪称量》给出的定量结论是什么?

结论:第一章用乘积极限法估计留存,首周顿悟组与非顿悟组的半年生存率相差 0.3069;同时提出核心反常识命题:在无人改变的异质性模型中,群组周流失风险八周内下降 0.5052,变平的曲线不等于会员变忠诚。

数据口径:书中参数为示意;全部计算由书中程序生成;本书不是法律意见,也不是个人化的财务、投资或经营建议。 书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。

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: 0.3069

weekKM ahaKM no-ahaKM alltrue allnaive all
10.99110.93880.95950.95920.9595
40.94230.77900.84350.84960.8480
80.89690.60320.71850.72890.7350
120.84310.55350.66720.68340.6925
260.71280.41330.53120.54650.6030
400.58480.32320.42740.43840.5645
true 26-week survival gap, aha minus no-aha: 0.3069
true 26-week churn 0.4535; naive count shows 0.3970
查看完整验算程序
# Chapter 1. p11: Kaplan-Meier survival of a paid community, by first week.
# Model: 2,000 members join uniformly over weeks 0 to 51; the analysis is
# run at week 52, so late joiners are censored. A member who reaches a
# first-week epiphany (share 0.40) has a constant weekly churn hazard of
# 0.012; one who does not has 0.060 for weeks 1 to 8, then 0.020.
# Seeded simulation; the true curves are computed in closed form.
# The naive column counts churned members over all members and so
# treats every censored member as a survivor. Parameters illustrative.
import random

rng = random.Random(20261112)
N, END, P_AHA = 2000, 52, 0.40


def hazard(aha, w):
    if aha:
        return 0.012
    return 0.060 if w <= 8 else 0.020


def true_surv(aha, w):
    s = 1.0
    for k in range(1, w + 1):
        s *= 1 - hazard(aha, k)
    return s


members = []
for _ in range(N):
    aha = rng.random() < P_AHA
    horizon = END - rng.randrange(0, END)
    t, dead = 0, False
    while t < horizon:
        t += 1
        if rng.random() < hazard(aha, t):
            dead = True
            break
    members.append((aha, t, dead))


def km(rows, last):
    s, out = 1.0, {}
    for w in range(1, last + 1):
        risk = len([r for r in rows if r[1] >= w])
        died = len([r for r in rows if r[1] == w and r[2]])
        if risk:
            s *= 1 - died / risk
        out[w] = s
    return out


WEEKS = (1, 4, 8, 12, 26, 40)
k_aha = km([m for m in members if m[0]], 40)
k_no = km([m for m in members if not m[0]], 40)
k_all = km(members, 40)
print("week  KM aha  KM no-aha  KM all  true all  naive all")
for w in WEEKS:
    tru = P_AHA * true_surv(True, w) + (1 - P_AHA) * true_surv(False, w)
    gone = len([m for m in members if m[2] and m[1] <= w])
    print("{:4d}  {:6.4f}  {:9.4f}  {:6.4f}  {:8.4f}  {:9.4f}".format(
        w, k_aha[w], k_no[w], k_all[w], tru, 1 - gone / N))
gap = true_surv(True, 26) - true_surv(False, 26)
print("true 26-week survival gap, aha minus no-aha: {:.4f}".format(gap))
tru26 = P_AHA * true_surv(True, 26) + (1 - P_AHA) * true_surv(False, 26)
gone26 = len([m for m in members if m[2] and m[1] <= 26]) / N
print("true 26-week churn {:.4f}; naive count shows {:.4f}".format(
    1 - tru26, gone26))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《高客单社群身份棘轮、可变奖励闯关上瘾与首周顿悟留存工程》第 1 章「首周分叉:乘积极限生存曲线、异质性的假象与顿悟率的价格」(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-148 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026148,
  author    = {Fan, Yafu},
  title     = {Identity Ratchets in High-Ticket Communities, Variable-Reward Quest Loops and First-Week Epiphany Retention Engineering},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论