关于「把降价与追单折回季前订量:有效残值与快速反应」,《季前订货临界分位母典》给出的定量结论是什么?

结论:第六章把降价期权与季中追单折回季前订量,早期销售与全季相关系数为 0.9 时,追单使利润提高 4.47 %。

数据口径:书中的商品、门店与需求参数均为示意;全部计算由书中程序生成;本书是非个人化经营建议。 第一章用边际算术推出临界分位,示意参数下只有毛利率恰为 0.3750 时最优订量才等于平均数,毛利八成的商品按平均数订会少赚 6.96 %;据施韦策与卡雄实验的已发表数据重建,受试者把订量向平均数拉了 0.644(高利润)与 0.787(低利润)的距离。

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: 0.9

single buy: q = 232.5, E[profit] = 10298.3
rbest q1E[profit]gain vs single
0.023210298.1-0.0000
0.522610345.8+0.0046
0.721410466.5+0.0163
0.919610758.1+0.0447

核心公式

\[q^{**} = \arg\max_q\; V_{10}(q) - c\,q,\]
查看完整验算程序
# Chapter 6. p62: Value of a second, in-season order (quick response).
# Season demand D ~ Normal(200, 60). Early sales give a standardised
# signal S with corr(S, D) = r, so D | S ~ N(200 + 60 r S,
# 60^2 (1 - r^2)). Price 100, outlet 15. First buy q1 at unit cost 40
# before the season; after S, top up to the conditional fractile
# target at unit cost 50 (express premium). q1 chosen by grid search
# on 20000 fixed-seed (S, D) pairs; the same demand draws serve every
# rule. Single buy = plain newsvendor. Values illustrative.
import math
import random
from statistics import NormalDist

rng = random.Random(20261112)
Z = NormalDist()
MU, SD, P, V, C1, C2, REPS = 200.0, 60.0, 100.0, 15.0, 40.0, 50.0, 20000
Z1 = [rng.gauss(0.0, 1.0) for _ in range(REPS)]
Z2 = [rng.gauss(0.0, 1.0) for _ in range(REPS)]


def profit(stock, d, c_part):
    sold = min(stock, d)
    return P * sold + V * (stock - sold) - c_part


def run(r):
    zt = Z.inv_cdf((P - C2) / (P - V))
    sc = SD * math.sqrt(1.0 - r * r)
    pairs = []
    for a, b in zip(Z1, Z2):
        s = r * a + math.sqrt(1.0 - r * r) * b
        d = max(0.0, MU + SD * a)
        pairs.append((MU + SD * r * s + sc * zt, d))
    best = None
    for q1 in range(100, 302, 2):
        vals = []
        for target, d in pairs:
            q2 = max(0.0, target - q1)
            vals.append(profit(q1 + q2, d, C1 * q1 + C2 * q2))
        m = math.fsum(vals) / REPS
        if best is None or m > best[1]:
            best = (q1, m)
    return best


q_single = MU + SD * Z.inv_cdf((P - C1) / (P - V))
single = math.fsum(profit(q_single, max(0.0, MU + SD * a), C1 * q_single)
                   for a in Z1) / REPS
print("single buy: q = {:.1f}, E[profit] = {:.1f}".format(q_single, single))
print("r     best q1  E[profit]  gain vs single")
for r in (0.0, 0.5, 0.7, 0.9):
    q1, m = run(r)
    print("{:.1f}   {:7d}  {:9.1f}  {:+.4f}".format(r, q1, m, m / single - 1))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《买手季前订货报童临界分位、尺码曲线狄利克雷先验与季末降价动态规划母典》第 6 章「把降价与追单折回季前订量:有效残值与快速反应」(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-211 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026211,
  author    = {Fan, Yafu},
  title     = {Newsvendor Critical Fractiles for Pre-Season Buying, Dirichlet Size-Curve Priors and Dynamic-Programming End-of-Season Markdowns},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论