关于「免邮门槛:分段期望毛利与营收陷阱」,《熟客唤醒与凑单闪购》给出的定量结论是什么?
结论:免邮门槛是一道价格,门槛两侧客人面对不同总价。分段期望毛利显示,营收最优门槛贴近篮子中位数,毛利最优门槛远高于它;按示意参数,前者是 RM 83,后者是 RM 170。凑单品毛利率越高、运费越低,最优门槛越低。全部免邮在示意参数下总是输给不免邮。用 Fraction 在经验分布上精确核对,可以避免舍入颠倒排序,也免去分布假设。
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: 170, 83
| mean intended basket RM 98.88; median RM 85.00 | ||||
| T | margin/visitor | revenue/visitor | P(order) | P(top-up) |
| 0 | 17.664 | 98.88 | 1.000 | 0.000 |
| 50 | 18.422 | 100.38 | 0.986 | 0.111 |
| 75 | 20.067 | 102.09 | 0.946 | 0.194 |
| 100 | 21.644 | 101.74 | 0.896 | 0.199 |
| 125 | 22.617 | 99.35 | 0.850 | 0.160 |
| 150 | 23.032 | 96.07 | 0.816 | 0.114 |
| 200 | 23.020 | 90.20 | 0.777 | 0.049 |
| 300 | 22.533 | 84.89 | 0.755 | 0.008 |
| 400 | 22.337 | 83.61 | 0.751 | 0.001 |
| margin-best T = RM 170: margin 23.109, revenue 93.47 | ||||
| revenue-best T = RM 83: margin 20.621, revenue 102.25 | ||||
| margin given up by the revenue-best T: 2.488 per visitor | ||||
| rule of thumb T = 1.1 x mean basket = RM 109: margin 22.068 |
查看完整验算程序
# Chapter 4. p41: expected gross margin per visitor against the free-
# shipping threshold T. Illustrative parameters, not a shop's data:
# intended basket B ~ lognormal, median RM 85, log-sd 0.55
# item margin rate MU 0.30; shipping cost to the shop S = RM 12
# below T the buyer is charged F = RM 12, and then either
# tops up with filler worth (T - B) + 4 with chance exp(-(T-B)/25),
# filler margin rate MUF 0.15 (accessories, tail stock), or
# orders as she is (paying F) with chance 1 - ALPHA, ALPHA = 0.25
# at or above T she orders B and the shop pays S.
# Expectation by quadrature over z (standard normal) on a fine grid.
import math
MED, SD, MU, MUF, S, F, ALPHA, DELTA, OVER = (
85.0, 0.55, 0.30, 0.15, 12.0, 12.0, 0.25, 25.0, 4.0)
Z = [i / 200.0 for i in range(-1200, 1201)]
WZ = [math.exp(-z * z / 2.0) / math.sqrt(2.0 * math.pi) / 200.0 for z in Z]
BASKETS = [MED * math.exp(SD * z) for z in Z]
def outcome(b, t, mu=MU, muf=MUF, ship=S):
"""(margin, revenue, P(order), P(top-up)) for one intended basket."""
if b >= t:
return mu * b - ship, b, 1.0, 0.0
d = t - b
tau = math.exp(-d / DELTA)
stay = (1.0 - tau) * (1.0 - ALPHA)
m = tau * (mu * b + muf * (d + OVER) - ship) + stay * (mu * b + F - ship)
rev = tau * (b + d + OVER) + stay * (b + F)
return m, rev, tau + stay, tau
def expect(t, **kw):
cols = [[], [], [], []]
for w, b in zip(WZ, BASKETS):
for j, v in enumerate(outcome(b, t, **kw)):
cols[j].append(w * v)
return [math.fsum(c) for c in cols]
def best_t(**kw):
grid = [float(t) for t in range(0, 401)]
return max(grid, key=lambda t: expect(t, **kw)[0])
if __name__ == "__main__":
mean_b = math.fsum(w * b for w, b in zip(WZ, BASKETS))
print("mean intended basket RM {:.2f}; median RM {:.2f}".format(
mean_b, MED))
print(" T margin/visitor revenue/visitor P(order) P(top-up)")
for t in (0, 50, 75, 100, 125, 150, 200, 300, 400):
m, r, po, pt = expect(float(t))
print("{:>5} {:7.3f} {:7.2f} {:.3f} {:.3f}"
.format(t, m, r, po, pt))
tm = best_t()
grid = [float(t) for t in range(0, 401)]
tr = max(grid, key=lambda t: expect(t)[1])
print("margin-best T = RM {:.0f}: margin {:.3f}, revenue {:.2f}".format(
tm, expect(tm)[0], expect(tm)[1]))
print("revenue-best T = RM {:.0f}: margin {:.3f}, revenue {:.2f}".format(
tr, expect(tr)[0], expect(tr)[1]))
print("margin given up by the revenue-best T: {:.3f} per visitor"
.format(expect(tm)[0] - expect(tr)[0]))
t_aov = round(mean_b * 1.1)
print("rule of thumb T = 1.1 x mean basket = RM {}: margin {:.3f}".format(
t_aov, expect(float(t_aov))[0]))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 4 章「免邮门槛:分段期望毛利与营收陷阱」(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}
}