关于「买到死模型:仍活跃概率与下一季期望单数」,《熟客唤醒与凑单闪购》给出的定量结论是什么?

结论:表里最醒目的是右下角。复购 12 次的顾客,沉默 8 周时仍活跃概率还有 0.583,沉默 14 周就跌到 0.126,沉默 20 周只剩 0.009。复购只有 1 次的顾客,沉默 20 周仍有 0.587,沉默 34 周还有 0.314。第二个表把这层意思压成一行:仍活跃概率跌破一半所需的沉默,复购 1 次要 26 周,复购 2 次要 21 周,复购 4 次要 16 周,复购 8 次只要 12 周,复购 12 次只要 9 周。

数据口径:示意参数下,面额 RM 6 的券只在间隔比 0.85 到 11.37 之间净值为正,刚买客与久睡客都是负值;不设对照组的核销记账会把增量高估 12.2 倍。 书中参数为示意,不对应 HOMIE-X 任何真实客户、订单或供货商,书中价格一律为示意;全部计算由书中程序生成。

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: 0.009, 0.126, 0.314, 0.583, 0.587, 12

P(alive) at the end of a 39-week window
weeks silent0 4 8 14 20 26 34
x1 0.754 0.731 0.704 0.653 0.587 0.496 0.314
x2 0.812 0.777 0.732 0.643 0.520 0.358 0.123
x4 0.872 0.820 0.742 0.567 0.333 0.124 0.010
x8 0.922 0.843 0.689 0.324 0.068 0.006 0.000
x12 0.944 0.838 0.583 0.126 0.009 0.000 0.000
silence at which P(alive) first drops below 0.5
x1: 26 weeks
x2: 21 weeks
x4: 16 weeks
x8: 12 weeks
x12: 9 weeks
mean purchase rate of a live buyer 0.0551 per week
mean drop-out chance per order a / (a + b)0.2463
查看完整验算程序
# Chapter 2. p21: BG/NBD probability that a buyer is still active.
# P(alive | x, t_x, T) = 1 / (1 + [x > 0] * a / (b + x - 1)
#                        * ((alpha + T) / (alpha + t_x)) ** (r + x))
# x = repeat orders in the calibration window, t_x = week of the last
# order, T = length of the window in weeks. Parameters are the published
# CDNOW estimates (r 0.243, alpha 4.414, a 0.793, b 2.426, weeks,
# T = 39, fitted on a sample of 2,357 buyers); they illustrate the shape
# only and are not a reseller's data.
import math

R, ALPHA, A, B, T = 0.243, 4.414, 0.793, 2.426, 39.0


def p_alive(x, tx, big_t=T):
    if x == 0:
        return 1.0
    ratio = (ALPHA + big_t) / (ALPHA + tx)
    return 1.0 / (1.0 + A / (B + x - 1.0) * ratio ** (R + x))


if __name__ == "__main__":
    print("P(alive) at the end of a 39-week window")
    print("weeks silent    0      4      8     14     20     26     34")
    for x in (1, 2, 4, 8, 12):
        row = []
        for silent in (0, 4, 8, 14, 20, 26, 34):
            row.append("{:.3f}".format(p_alive(x, T - silent)))
        print("x = {:>2}      ".format(x) + "  ".join(row))
    print("silence at which P(alive) first drops below 0.5")
    for x in (1, 2, 4, 8, 12):
        s = 0
        while s < 39 and p_alive(x, T - s) >= 0.5:
            s += 1
        txt = "never inside 39 weeks" if s == 39 else "{} weeks".format(s)
        print("  x = {:>2}: {}".format(x, txt))
    mean_rate = R / ALPHA
    print("mean purchase rate of a live buyer {:.4f} per week".format(
        mean_rate))
    print("mean drop-out chance per order a / (a + b) = {:.4f}".format(
        A / (A + B)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《私域代购熟客复购唤醒、满额凑单毛利配比与限时清仓脉冲闪购母典》第 2 章「买到死模型:仍活跃概率与下一季期望单数」(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-353 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026353,
  author    = {Fan, Yafu},
  title     = {Repeat-Buyer Reactivation in Private-Channel Commerce, Threshold Bundle Margin Design and Flash-Clearance Dynamics},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论