关于「耐心的形状:指数、威布尔与对数正态的风险率,以及删失下的耐心估计」,《私域等待、转人工门限与客诉阻断》给出的定量结论是什么?
结论:第一章说明,同样的平均耐心下,风险率可以上升、下降或先升后降,指数耐心下等过任何时刻的买手在下 30 秒离开的概率都是 0.221;下降的总体风险率可能只是急性子先走光,聊天记录的删失又会让直接平均出错。
数据口径:书中参数为示意;全部计算由书中程序生成;本书不是法律、财务或经营上的个人化建议。 书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。
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: 0.221, 30
| still waiting S(t), mean patience 120 s in every law | ||||
| t (s) | EXP | W1.6 | W0.6 | LN1.0 |
| 0 | 1.000 | 1.000 | 1.000 | 1.000 |
| 15 | 0.882 | 0.970 | 0.693 | 0.943 |
| 30 | 0.779 | 0.913 | 0.573 | 0.812 |
| 60 | 0.607 | 0.758 | 0.430 | 0.577 |
| 120 | 0.368 | 0.432 | 0.279 | 0.309 |
| 240 | 0.135 | 0.078 | 0.144 | 0.116 |
| 480 | 0.018 | 0.000 | 0.053 | 0.030 |
| P(leave within next 30 s | waited t) | ||||
| 0 | 0.221 | 0.087 | 0.427 | 0.188 |
| 15 | 0.221 | 0.135 | 0.290 | 0.274 |
| 30 | 0.221 | 0.169 | 0.249 | 0.290 |
| 60 | 0.221 | 0.223 | 0.207 | 0.279 |
| 120 | 0.221 | 0.303 | 0.167 | 0.239 |
| 240 | 0.221 | 0.410 | 0.132 | 0.184 |
| 480 | 0.221 | 0.544 | 0.103 | 0.130 |
| median | 83.2 | 106.4 | 43.3 | 72.8 |
查看完整验算程序
# Chapter 1. p11: three patience laws with the same mean, three hazards.
# Model (illustrative parameters, not data): a buyer who sends a private
# message waits a random "patience" T seconds and leaves if no reply has
# come. All four laws below have mean patience 120 s:
# EXP exponential, constant hazard
# W1.6 Weibull shape 1.6, rising hazard ("the longer, the likelier
# to leave")
# W0.6 Weibull shape 0.6, falling hazard ("survive the start, stay")
# LN1.0 lognormal, log-sd 1.0, hazard rises then falls
# For each law the program prints the share still waiting S(t) and the
# chance of leaving within the next 30 s for a buyer who has waited t.
import math
MEAN = 120.0
def weib(k):
lam = MEAN / math.gamma(1.0 + 1.0 / k)
return lambda t: math.exp(-(t / lam) ** k)
def lognorm(sd):
mu = math.log(MEAN) - sd * sd / 2.0
def s(t):
if t <= 0.0:
return 1.0
z = (math.log(t) - mu) / sd
return 0.5 * math.erfc(z / math.sqrt(2.0))
return s
LAWS = (("EXP", lambda t: math.exp(-t / MEAN)), ("W1.6", weib(1.6)),
("W0.6", weib(0.6)), ("LN1.0", lognorm(1.0)))
TS = (0, 15, 30, 60, 120, 240, 480)
def median(s):
lo, hi = 0.0, 1.0e5
for _ in range(80):
mid = (lo + hi) / 2.0
if s(mid) > 0.5:
lo = mid
else:
hi = mid
return lo
print("still waiting S(t), mean patience {:.0f} s in every law".format(
MEAN))
print("t (s) " + "".join("{:>8}".format(n) for n, _ in LAWS))
for t in TS:
print("{:5d} ".format(t) + "".join(
"{:8.3f}".format(s(t)) for _, s in LAWS))
print("P(leave within next 30 s | waited t)")
for t in TS:
print("{:5d} ".format(t) + "".join(
"{:8.3f}".format(1.0 - s(t + 30) / s(t)) for _, s in LAWS))
print("median " + "".join("{:8.1f}".format(median(s)) for _, s in LAWS))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《私域对话队列耐心衰减分布、人机协同红黄绿灯移交门限与客诉升级马尔可夫阻断》第 1 章「耐心的形状:指数、威布尔与对数正态的风险率,以及删失下的耐心估计」(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-175 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026175,
author = {Fan, Yafu},
title = {Patience-Decay Distributions in Private-Message Queues, Red-Amber-Green Human-AI Handoff Thresholds and Markov Blocking of Complaint Escalation},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}