关于「唤醒券增量账:存活概率、必回客漏损与发券区间」,《熟客唤醒与凑单闪购》给出的定量结论是什么?
结论:唤醒券有四个去处,只有一个是增量。刚下过单的熟客必回概率最高,券多半是白送;久睡熟客仍活跃概率太低,券多半是空投。增量价值在中段最高,按示意参数,面额 RM 6 的券在间隔比 0.85 到 11.37 之间才划算,节奏越清楚的熟客区间越窄。核销记账会把增量高估十倍以上,必须留一组不发券。
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。
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: 0.85, 11.37
| g | P_alive | anyway b | lift u | gain | leak | net RM |
|---|---|---|---|---|---|---|
| 0.0 | 0.850 | 0.283 | 0.108 | 1.487 | 1.442 | -0.205 |
| 0.5 | 0.803 | 0.264 | 0.110 | 1.449 | 1.272 | -0.074 |
| 1.0 | 0.749 | 0.248 | 0.113 | 1.393 | 1.115 | 0.028 |
| 2.0 | 0.633 | 0.221 | 0.117 | 1.242 | 0.838 | 0.154 |
| 3.0 | 0.515 | 0.199 | 0.120 | 1.068 | 0.615 | 0.203 |
| 4.0 | 0.408 | 0.181 | 0.123 | 0.897 | 0.443 | 0.204 |
| 6.0 | 0.247 | 0.153 | 0.127 | 0.622 | 0.227 | 0.145 |
| 9.0 | 0.117 | 0.125 | 0.131 | 0.386 | 0.087 | 0.049 |
| 12.0 | 0.059 | 0.105 | 0.134 | 0.278 | 0.037 | -0.010 |
| 16.0 | 0.027 | 0.087 | 0.137 | 0.215 | 0.014 | -0.049 |
| coupon RM 6: net > 0 for g in [0.85, 11.37], best g 3.49, 0.208 RM per send | ||||||
| coupon RM 4: net > 0 for g in [0.00, 14.34], best g 1.68 | ||||||
| coupon RM 8: no gap ratio pays | ||||||
| KAPPA | 2.0: net > 0 for g in [0.29, 20.00], best 3.64 | |||||
| KAPPA | 7.0: net > 0 for g in [1.83, 6.89], best 3.64 | |||||
| KAPPA 11.0: no gap ratio pays |
查看完整验算程序
# Chapter 3. p31: incremental value of one reactivation coupon by gap.
# Gap ratio g = days since the last order / the buyer's shrunk personal
# median gap (p12). A live buyer orders as a Poisson process whose rate
# lam is uncertain: lam ~ Gamma(KAPPA, KAPPA / LAM0), LAM0 = ln 2 (median
# gap 1), KAPPA = n + K0 = 4 (one own gap plus the K0 = 3 of p12).
# After an order she stays active with probability PI = 0.85.
# L(g) = (1 + g LAM0 / KAPPA) ** -KAPPA P(no order by g | live)
# P_alive = PI L / (PI L + 1 - PI)
# b = 1 - ((KAPPA/LAM0 + g) / (KAPPA/LAM0 + g + W)) ** KAPPA
# (a live buyer orders in the next window anyway)
# u = THETA (1 - b) coupon converts some who would not
# net = P_alive (u (m - c) - b c) + (1 - P_alive) RHO (m - c) - s
# Illustrative: window W = 0.5 median, THETA 0.15, a lapsed buyer comes
# back on a coupon with RHO 0.01, margin m RM 22, coupon c RM 6, and s =
# RM 0.25 per touch (message credits and opt-out risk, illustrative).
import math
PI, THETA, RHO, W, SEND = 0.85, 0.15, 0.01, 0.5, 0.25
KAPPA, LAM0 = 4.0, math.log(2.0)
M, C = 22.0, 6.0
def parts(g, m=M, c=C, kappa=KAPPA):
beta = kappa / LAM0
lg = (beta / (beta + g)) ** kappa
pa = PI * lg / (PI * lg + 1.0 - PI)
b = 1.0 - ((beta + g) / (beta + g + W)) ** kappa
u = THETA * (1.0 - b)
gain = pa * u * (m - c) + (1.0 - pa) * RHO * (m - c)
leak = pa * b * c
return pa, b, u, gain, leak, gain - leak - SEND
def band(c=C, kappa=KAPPA):
grid = [i / 100.0 for i in range(0, 2001)]
pos = [g for g in grid if parts(g, c=c, kappa=kappa)[5] > 0.0]
best = max(grid, key=lambda g: parts(g, c=c, kappa=kappa)[5])
return pos, best
if __name__ == "__main__":
print(" g P_alive anyway b lift u gain leak net RM")
for g in (0.0, 0.5, 1.0, 2.0, 3.0, 4.0, 6.0, 9.0, 12.0, 16.0):
pa, b, u, gain, leak, net = parts(g)
print("{:>4} {:.3f} {:.3f} {:.3f} {:.3f} {:.3f} {:6.3f}"
.format(g, pa, b, u, gain, leak, net))
pos, best = band()
print("coupon RM 6: net > 0 for g in [{:.2f}, {:.2f}], best g {:.2f},"
" {:.3f} RM per send".format(pos[0], pos[-1], best,
parts(best)[5]))
for c in (4.0, 8.0):
pos, best = band(c=c)
if pos:
print("coupon RM {:.0f}: net > 0 for g in [{:.2f}, {:.2f}],"
" best g {:.2f}".format(c, pos[0], pos[-1], best))
else:
print("coupon RM {:.0f}: no gap ratio pays".format(c))
for kappa in (2.0, 7.0, 11.0):
pos, best = band(kappa=kappa)
if not pos:
print("KAPPA {:>4.1f}: no gap ratio pays".format(kappa))
continue
print("KAPPA {:>4.1f}: net > 0 for g in [{:.2f}, {:.2f}], best {:.2f}"
.format(kappa, pos[0], pos[-1], best))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 3 章「唤醒券增量账:存活概率、必回客漏损与发券区间」(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}
}