关于「意向衰减:十五分钟、二十四小时与七十二小时三波召回」,《弃单召回与升单棘轮》给出的定量结论是什么?
结论:B 段只移动第一波。放在一分钟时,第一波只挽回 0.00609,净值 +0.0677,三波合计 0.7938;五分钟时挽回 0.02284,合计 1.4113;十五分钟时 0.03732,合计 1.9419;二十五分钟时 0.03869,净值 +1.2823,合计 1.9941,是表中最高;一小时时回落到 0.03301,合计 1.7972;六小时时只有 0.01412,合计 1.1494。曲线在十五到二十五分钟之间达到顶点,两侧都下降:太早,新意不足、自返者多、退订高;太晚,冲动型访客的意向已经散了。
数据口径:按本书的示意模型,第一波召回放在一分钟时只挽回同意离开者的 0.00609,放在十五分钟时挽回 0.03732;把三波压在一小时之内,每位离开者净值 RM 1.1790,按十五分钟、二十四小时、七十二小时发送时为 RM 1.9419;第五波的净值已是 -0.0121。 书中网站被当作付费报告产品讨论,不宣称任何预测准确性,不用恐惧招揽;所有访客与订单数据均为合成或示意。
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: 0.00609, 0.01412, 0.02284, 0.03301, 0.03732, 0.03869
| value of one recovered buyer: RM 35.92 | |
| A. 15 min / 24 h / 72 h, then two extra waves at 7 and 14 days | |
| wave time recovered unsubscribed net RM per leaver | |
| 1 | 15 min 0.03732 0.01389 +1.2293 |
| 2 | 24 h 0.01517 0.00505 +0.5047 |
| 3 | 72 h 0.00668 0.00399 +0.2080 |
| 4 | 7 d 0.00220 0.00445 +0.0432 |
| 5 | 14 d 0.00084 0.00529 -0.0121 |
| first three waves recover 0.05917; shares 0.631, 0.256, 0.113 | |
| net of three waves RM 1.9419; of five RM 1.9731 | |
| B. moving only the first wave (waves 2 and 3 stay at 24 h and 72 h) | |
| first wave wave-1 recovered wave-1 net RM net of three RM | |
| 1 min | 0.00609 +0.0677 0.7938 |
| 5 min | 0.02284 +0.6921 1.4113 |
| 15 min | 0.03732 +1.2293 1.9419 |
| 25 min | 0.03869 +1.2823 1.9941 |
| 1 h | 0.03301 +1.0834 1.7972 |
| 6 h | 0.01412 +0.4352 1.1494 |
| 'as soon as possible' 1 min / 10 min / 1 h: net RM 1.1790, unsubscribed 0.06416 |
查看完整验算程序
# Chapter 2. p23: marginal recovery and net value of each recall wave.
# Illustrative model per consented leaver (all parameters synthetic):
# self-returners, share 0.03, pay anyway (lognormal return, median
# 25 min, log-sd 1.6); a message to them earns nothing, costs risk.
# the rest split into impulse visitors (0.6) whose intent decays as a
# power law (1 + t/30 min)^-0.45 and deliberators (0.4) whose
# readiness rises over 12 h and fades over 3 days.
# a message read within minutes repeats what is still on screen:
# novelty factor 1 - exp(-t / 8 min).
# time-of-day factor 0.6 + 0.4 cos(2 pi t / 24 h); fatigue 0.75 per
# earlier message.
# unsubscribe risk 0.002 x (1 + 0.5 per earlier message) x (1 +
# 3 exp(-t / 5 min) + 6 exp(-gap / 12 h)); complaints = 0.1 x that.
# value of a recovered buyer RM 35.93 (RM 19.90 plus a 0.10 chance of
# the RM 179.10 upgrade, after 0.05 fees and refunds); a lost
# subscriber costs RM 4.00 of list value; a complaint RM 40.00.
import math
from statistics import NormalDist
ND = NormalDist()
P = dict(rho=0.03, wa=0.6, tau=30.0, beta=0.45, ka=0.09, theta=8.0, kb=0.05,
ready=720.0, fade=4320.0, fat=0.75, u0=0.002, a1=3.0, phi=5.0, a2=6.0,
psi=720.0, g=0.5, cpu=0.1, list_value=4.0, complaint=40.0)
MARGIN = 0.95 * (19.90 + 0.10 * (199.0 - 19.90))
def diurnal(t):
return 0.6 + 0.4 * math.cos(2.0 * math.pi * t / 1440.0)
def run(times, p=P):
a = b = srs = 1.0
rows = []
prev = None
for n, t in enumerate(times, 1):
gap = t if prev is None else t - prev
d = diurnal(t) * p["fat"] ** (n - 1)
qa = p["ka"] * (1 + t / p["tau"]) ** -p["beta"] * (1 - math.exp(-t / p["theta"])) * d
qb = p["kb"] * (1 - math.exp(-t / p["ready"])) * math.exp(-t / p["fade"]) * d
u = p["u0"] * (1 + (n - 1) * p["g"]) * (
1 + p["a1"] * math.exp(-t / p["phi"]) + p["a2"] * math.exp(-gap / p["psi"]))
left = 1.0 - ND.cdf(math.log(t / 25.0) / 1.6)
rest = 1.0 - p["rho"]
rec = rest * (p["wa"] * a * qa + (1 - p["wa"]) * b * qb)
reach = rest * (p["wa"] * a + (1 - p["wa"]) * b) + p["rho"] * srs * left
uns = reach * u
val = MARGIN * rec - p["list_value"] * uns - p["complaint"] * p["cpu"] * uns
rows.append((n, t, rec, uns, val))
a, b, srs = a * (1 - qa) * (1 - u), b * (1 - qb) * (1 - u), srs * (1 - u)
prev = t
return rows
def label(t):
if t < 60:
return "{:g} min".format(t)
if t <= 4320:
return "{:g} h".format(t / 60)
return "{:g} d".format(t / 1440)
if __name__ == "__main__":
print("value of one recovered buyer: RM {:.2f}".format(MARGIN))
print("A. 15 min / 24 h / 72 h, then two extra waves at 7 and 14 days")
print(" wave time recovered unsubscribed net RM per leaver")
rows = run([15, 1440, 4320, 10080, 20160])
for n, t, rec, uns, val in rows:
print(" {} {:<7} {:.5f} {:.5f} {:+.4f}".format(n, label(t), rec, uns, val))
three = rows[:3]
tot = math.fsum(r[2] for r in three)
print(" first three waves recover {:.5f}; shares {}".format(
tot, ", ".join("{:.3f}".format(r[2] / tot) for r in three)))
print(" net of three waves RM {:.4f}; of five RM {:.4f}".format(
math.fsum(r[4] for r in three), math.fsum(r[4] for r in rows)))
print("B. moving only the first wave (waves 2 and 3 stay at 24 h and 72 h)")
print(" first wave wave-1 recovered wave-1 net RM net of three RM")
for t1 in (1, 5, 15, 25, 60, 360):
r = run([t1, 1440, 4320])
print(" {:<11} {:.5f} {:+.4f} {:.4f}".format(
label(t1), r[0][2], r[0][4], math.fsum(x[4] for x in r)))
r = run([1, 10, 60])
print(" 'as soon as possible' 1 min / 10 min / 1 h: net RM {:.4f}, unsubscribed {:.5f}".format(
math.fsum(x[4] for x in r), math.fsum(x[3] for x in r)))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《诊断收费站弃单挽回时序、个性化悬念召回与高客单升单棘轮母典》第 2 章「意向衰减:十五分钟、二十四小时与七十二小时三波召回」(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-349 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026349,
author = {Fan, Yafu},
title = {Paywall Abandonment Recovery Timing, Personalised Curiosity Recall and High-Ticket Upsell Ratchets},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}