关于「进箱物理:认证、投诉率与信誉后验」,《广告库反查与进箱物理》给出的定量结论是什么?

结论:第三、四章把发信信誉写成投诉率的贝塔后验:发信量越过匹配人群,第 30 天进箱率可从 0.974 跌到 0.054;预热是读后验的速率控制器,一键退订是压低投诉的物理装置。

数据口径:公开广告库里仍在投的广告是右删失数据,须用卡普兰与迈耶的乘积极限估计去读;「仍在投且已投满 30 天」这条规则在合成广告库里把持续投放者的比例从 0.313 抬到 0.902。 书中参数为示意,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: 0.054, 0.974, 30

per day complaint % inbox d1 d7 d14 d30 replies/30d
600.0600.8890.958 0.971 0.97452.0
1200.0600.8890.958 0.971 0.974103.9
1500.1680.8890.703 0.643 0.62874.0
2000.2760.8890.397 0.298 0.27641.2
4000.4380.8890.158 0.098 0.08721.0
8000.5190.8890.105 0.062 0.05418.9
more mail past the matched pool lowers total replies
查看完整验算程序
# Chapter 3. p33: the inbox cliff when volume outruns relevance.
# Expected-value model (illustrative parameters, no randomness):
#   M = 120 well-matched public business contacts per day; any extra
#   volume reaches poorly matched contacts.
#   complaint probability: matched 0.0006, poorly matched 0.006
#   reply probability if inboxed: matched 0.03, poorly matched 0.002
#   reputation r = EWMA (weight 0.2) of the daily complaint rate
#   inbox placement p(r) = 1 / (1 + (r / 0.002) ** 3)
# Complaints can only come from mail that reached the inbox.
import math

M = 120
Q_REL, Q_IRR = 0.0006, 0.006
R_REL, R_IRR = 0.03, 0.002


def place(r):
    return 1.0 / (1.0 + (r / 0.002) ** 3)


def simulate(v, days=30):
    r = 0.001
    good, bad = min(v, M), max(0, v - M)
    q = (Q_REL * good + Q_IRR * bad) / v
    path, replies = [], []
    for _ in range(days):
        p = place(r)
        replies.append(p * (R_REL * good + R_IRR * bad))
        path.append(p)
        r = 0.8 * r + 0.2 * q
    return path, replies, q


if __name__ == "__main__":
    print(" per day  complaint %  inbox d1  d7    d14   d30  replies/30d")
    for v in (60, 120, 150, 200, 400, 800):
        path, rep, q = simulate(v)
        print(" {:>7}  {:>11.3f}  {:>8.3f}  {:.3f} {:.3f} {:.3f}  {:>11.1f}"
              .format(v, 100 * q, path[0], path[6], path[13], path[29],
                      math.fsum(rep)))
    print("more mail past the matched pool lowers total replies")

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《公开广告库金主反查、B2B冷触达进箱物理与高客单暗投签单母典》第 3 章「进箱物理:认证、投诉率与信誉后验」(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-347 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026347,
  author    = {Fan, Yafu},
  title     = {Active Ad-Library Prospecting, B2B Outbound Deliverability Physics and High-Ticket Whitelist Deal Closing},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论