关于「客诉升级马尔可夫链:吸收概率与最划算的阻断点」,《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:第五章把客诉升级写成吸收马尔可夫链,一次咨询最终走向公开曝光的概率是 0.01469,上游的首答阻断每花一元省下 24.18 元,但其单价超过 2.57 元时应改为不满介入。

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。

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: 2.57, 24.18

blockexposurescomplaintspendtotal costsaved per
rounds(yuan)(yuan)yuan spent
none146.9597.00530149
B-I73.1297.11057627439124.18
B-D72.7295.7283882909818.42
B-P102.1597.0358214316422.75
B-I keeps the larger net saving up to 2.57 yuan per inquiry round
查看完整验算程序
# Chapter 5. p52: where blocking the escalation chain pays the most.
# Model (illustrative parameters): the chain of p51.  Three blocks:
#   B-I  a complete first reply, 1 yuan per inquiry round, halves I->D
#   B-D  human apology and make-good, 15 yuan per dissatisfied round,
#        halves D->P and D->E
#   B-P  fast settlement of a platform complaint, 60 yuan per complaint
#        round, halves P->E
# Moved probability goes to "resolved".  Each exposure is charged
# X_EXP yuan and each complaint round X_CMP yuan (illustrative).  All
# figures are per 10000 inquiries; exact fractions, printed as floats.
from fractions import Fraction as F

X_EXP, X_CMP, PER = 3000, 150, 10000
T = ("I", "D", "P")
#          I        D         P         E          R
ROWS = {"I": (F(5, 100), F(15, 100), F(0), F(0), F(80, 100)),
        "D": (F(5, 100), F(15, 100), F(25, 100), F(3, 100), F(52, 100)),
        "P": (F(0), F(5, 100), F(20, 100), F(15, 100), F(60, 100))}


def inverse(m):
    n = len(m)
    a = [list(r) + [F(int(i == j)) for j in range(n)]
         for i, r in enumerate(m)]
    for c in range(n):
        p = next(r for r in range(c, n) if a[r][c] != 0)
        a[c], a[p] = a[p], a[c]
        piv = a[c][c]
        a[c] = [x / piv for x in a[c]]
        for r in range(n):
            if r != c and a[r][c] != 0:
                f = a[r][c]
                a[r] = [x - f * y for x, y in zip(a[r], a[c])]
    return [r[n:] for r in a]


def solve(rows):
    q = [[rows[i][j] for j in range(3)] for i in T]
    b = [[rows[i][3], rows[i][4]] for i in T]
    n = inverse([[F(int(i == j)) - q[i][j] for j in range(3)]
                 for i in range(3)])
    nb = [[n[i][0] * b[0][k] + n[i][1] * b[1][k] + n[i][2] * b[2][k]
           for k in range(2)] for i in range(3)]
    return n, nb


def halve(rows, state, cols):
    r = list(rows[state])
    for c in cols:
        r[4] += r[c] / 2
        r[c] = r[c] / 2
    out = dict(rows)
    out[state] = tuple(r)
    return out


def report(name, rows, unit, state_ix):
    n, nb = solve(rows)
    exp_, cmp_ = PER * nb[0][0], PER * n[0][2]
    spend = PER * unit * n[0][state_ix] if unit else F(0)
    total = exp_ * X_EXP + cmp_ * X_CMP + spend
    return name, exp_, cmp_, spend, total


cases = [report("none", ROWS, 0, 0),
         report("B-I", halve(ROWS, "I", (1,)), 1, 0),
         report("B-D", halve(ROWS, "D", (2, 3)), 15, 1),
         report("B-P", halve(ROWS, "P", (3,)), 60, 2)]
base = cases[0][4]
print("block  exposures  complaint  spend   total cost  saved per")
print("                  rounds     (yuan)  (yuan)      yuan spent")
for name, e, c, s, t in cases:
    ratio = "" if s == 0 else "{:10.2f}".format(float((base - t) / s))
    print("{:5}  {:9.1f}  {:9.1f}  {:6.0f}  {:10.0f}  {}".format(
        name, float(e), float(c), float(s), float(t), ratio))
# at 1 yuan per round, the B-I spend equals its count of inquiry rounds
gross_i = base - (cases[1][4] - cases[1][3])
net_d = base - cases[2][4]
price = (gross_i - net_d) / cases[1][3]
print("B-I keeps the larger net saving up to {:.2f} yuan per inquiry"
      " round".format(float(price)))

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 5 章「客诉升级马尔可夫链:吸收概率与最划算的阻断点」(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}
}

本书其他章节结论