关于「实证闭环:首周脉冲排程、切换日判定与预注册检验」,《新品冷启动的广告脉冲与口碑接棒》给出的定量结论是什么?
结论:第八章给出首周脉冲排程、切换日判定、小工具「冷启动投放切换点计算器」与预注册方案(每组 20 个单位时模拟检验力 0.840)。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。
数据口径:本书用巴斯扩散加广告控制的最优控制模型检验这一做法:广告以加法、线性方式作用于创新项并有每日上限时,庞特里亚金极大值原理与贝尔曼动态规划给出同一个答案,开头打满上限、在开关函数变负的那天一次停投,把接力棒交给口碑;示意参数下同额预算前六天打满比六十天均匀多赚 13.4 %。 书中的市场、毛利、扩散系数、广告效果、上限与产能均为示意参数;全部计算由书中程序生成;本书是非个人化经营建议,紧迫感只来自真实库存与真实截止。
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: 0.840, 20
| pilot: mean gain 0.1007 of margin per buyer, pooled sd 0.1115 | |
| standardized effect d | 0.903 |
| units per arm for 80 % power, two-sided 5 % | 20 |
| simulated power with 20 per arm | 0.840 |
| simulated power with 40 per arm | 0.981 |
查看完整验算程序
# Chapter 8. p83: How many launches does a fair test need?
# Units are separate launches or separate regional markets, randomized
# to a front pulse or to an even spread of the same budget. Each unit
# draws its own market size (lognormal, sd 0.3 on the log scale around
# 20000), q ~ U(0.05, 0.11), beta ~ U(2e-6, 6e-6), p = 0.001; budget
# 1.5 yuan and daily cap 0.25 yuan per potential buyer. The outcome is
# 60-day profit per potential buyer divided by the margin, plus
# measurement noise with sd 0.03 (all illustrative). Sample size from
# the normal approximation, then checked by simulation (fixed seed).
import math
import random
import statistics
MARGIN, P, T = 30.0, 0.001, 60
rng = random.Random(20261113)
Z = statistics.NormalDist()
def outcome(front):
m = 20000 * math.exp(rng.gauss(0, 0.3))
q = rng.uniform(0.05, 0.11)
b = rng.uniform(2e-6, 6e-6)
cap, budget = 0.25 * m, 1.5 * m
n = int(budget // cap)
us = [cap] * n + [0.0] * (T - n) if front else [budget / T] * T
x, cash = 0.0, []
for u in us:
x2 = x + (P + b * u * 20000 / m + q * x) * (1 - x)
cash.append(MARGIN * m * (x2 - x) - u)
x = x2
return math.fsum(cash) / (MARGIN * m) + rng.gauss(0, 0.03)
pilot_f = [outcome(True) for _ in range(2000)]
pilot_e = [outcome(False) for _ in range(2000)]
diff = statistics.fmean(pilot_f) - statistics.fmean(pilot_e)
sd = math.sqrt((statistics.variance(pilot_f)
+ statistics.variance(pilot_e)) / 2)
d = diff / sd
za, zb = Z.inv_cdf(0.975), Z.inv_cdf(0.80)
n = math.ceil(2 * (za + zb) ** 2 / d ** 2)
print("pilot: mean gain %.4f of margin per buyer, pooled sd %.4f"
% (diff, sd))
print("standardized effect d = %.3f" % d)
print("units per arm for 80 %% power, two-sided 5 %%: %d" % n)
for k in (n, 2 * n):
hits = 0
for _ in range(1000):
f = [outcome(True) for _ in range(k)]
e = [outcome(False) for _ in range(k)]
se = math.sqrt(statistics.variance(f) / k
+ statistics.variance(e) / k)
if (statistics.fmean(f) - statistics.fmean(e)) / se > za:
hits += 1
print("simulated power with %d per arm: %.3f" % (k, hits / 1000))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《庞特里亚金极大值原理、新品冷启动广告脉冲烧钱速率与口碑接棒最优控制母典》第 8 章「实证闭环:首周脉冲排程、切换日判定与预注册检验」(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-241 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026241,
author = {Fan, Yafu},
title = {Pontryagin's Maximum Principle at Product Launch: Optimal Ad-Pulse Burn Rates and the Hand-Over to Word of Mouth},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}