关于「范氏价格敏感度:四个问题、四个交点与它们管不到的地方」,《冷启动测价与贝叶斯调价》给出的定量结论是什么?
结论:范氏量表用四个开放题换来四条累积曲线,在栖桐工作室的虚构样本里,约定 A 下的四个交点是边际低价点 95.8 元、最优价格点 145.0 元、无差异价格点 147.6 元、边际高价点 212.1 元。约定 B 会把区间从 116.3 元宽压到 53.7 元宽,拿到报告先问约定。自举区间显示最优价格点最不稳,宽 17.9 元;样本量翻两番,区间宽度大约减半,有效问卷不少于 200 份。NMS 校准把边界接到收入上:试购率最高的 105 元不是收入最高的 140 元,校准权重改变的是销量的高低,不是价格的排名。锚定、混合人群与零购买意愿是三处已知的缺口。
数据口径:价格展示与分群定价的法规只写一般框架,不是法律或税务意见。 书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它;原典考辨取《管子·国蓄》的轻重敛散与「准平」。
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: 116.3, 145.0, 147.6, 212.1, 53.7, 95.8
| respondents after cleaning 277; distinct answered prices 81 | ||||||
| price | too_cheap | cheap | not_cheap | expensive | not_exp | too_exp |
| 40 | 0.989 | 1.000 | 0.000 | 0.000 | 1.000 | 0.000 |
| 70 | 0.621 | 0.982 | 0.018 | 0.000 | 1.000 | 0.000 |
| 100 | 0.188 | 0.736 | 0.264 | 0.014 | 0.986 | 0.000 |
| 130 | 0.040 | 0.347 | 0.653 | 0.126 | 0.874 | 0.011 |
| 160 | 0.000 | 0.152 | 0.848 | 0.332 | 0.668 | 0.065 |
| 190 | 0.000 | 0.040 | 0.960 | 0.592 | 0.408 | 0.184 |
| 220 | 0.000 | 0.014 | 0.986 | 0.773 | 0.227 | 0.361 |
| 250 | 0.000 | 0.004 | 0.996 | 0.881 | 0.119 | 0.523 |
| 280 | 0.000 | 0.000 | 1.000 | 0.942 | 0.058 | 0.675 |
| 310 | 0.000 | 0.000 | 1.000 | 0.971 | 0.029 | 0.780 |
| 340 | 0.000 | 0.000 | 1.000 | 0.989 | 0.011 | 0.859 |
| convention A: PMC 95.8 OPP 145.0 IPP 147.6 PME 212.1 range width 116.3 | ||||||
| convention B: PMC 117.6 OPP 145.0 IPP 147.6 PME 171.4 range width 53.7 | ||||||
| joined curves: at OPP too cheap = too expensive = 0.014; at IPP cheap = expensive = 0.228 | ||||||
| at PMC too cheap = not cheap = 0.209; at PME too expensive = not expensive = 0.278 | ||||||
| respondents for whom OPP is outside their own (too cheap, too expensive) band: 8 of 277 | ||||||
查看完整验算程序
# Chapter 2. p21: Cumulative Van Westendorp curves and the four intersections by linear interpolation.
# Same fictional sample as p20 (seed 13302, 300 respondents, cleaned). At each price p:
# TC(p) share whose too-cheap answer >= p (falls as p rises)
# CH(p) share whose cheap answer >= p (falls)
# EX(p) share whose expensive answer <= p (rises)
# TE(p) share whose too-expensive answer <= p (rises)
# The curves are evaluated at every price that anyone answered and joined by straight lines; a crossing
# is the root of the difference on the first segment where its sign changes.
# Convention A (with "not cheap" = 1 - CH and "not expensive" = 1 - EX):
# PMC = TC x (1 - CH), OPP = TC x TE, IPP = CH x EX, PME = TE x (1 - EX)
# Convention B (common in survey software): PMC = TC x EX, PME = TE x CH; OPP and IPP unchanged.
import bisect
import math
import random
SEED, N, CENTER, SIGMA = 13302, 300, 150.0, 0.28
def r5(x):
return max(5, 5 * round(x / 5))
def survey(n=N, seed=SEED):
rng = random.Random(seed)
rows = []
for _ in range(n):
m = math.exp(rng.gauss(math.log(CENTER), SIGMA))
tc, ch = r5(m * rng.uniform(0.40, 0.60)), r5(m * rng.uniform(0.70, 0.88))
ex, te = r5(m * rng.uniform(1.12, 1.35)), r5(m * rng.uniform(1.45, 1.95))
if rng.random() < 0.07:
ch, ex = ex, ch
rows.append((tc, ch, ex, te))
return [r for r in rows if r[0] < r[1] < r[2] < r[3]]
def curves(rows):
cols = [sorted(r[j] for r in rows) for j in range(4)]
n = len(rows)
knots = sorted(set(v for c in cols for v in c))
def ge(c, p):
return (n - bisect.bisect_left(c, p)) / n
def le(c, p):
return bisect.bisect_right(c, p) / n
tc = [ge(cols[0], p) for p in knots]
ch = [ge(cols[1], p) for p in knots]
ex = [le(cols[2], p) for p in knots]
te = [le(cols[3], p) for p in knots]
return knots, tc, ch, ex, te
def cross(knots, f, g):
d = [a - b for a, b in zip(f, g)]
for i in range(1, len(knots)):
if d[i - 1] > 0 >= d[i] or d[i - 1] < 0 <= d[i]:
if d[i] == 0:
return float(knots[i])
t = d[i - 1] / (d[i - 1] - d[i])
return knots[i - 1] + t * (knots[i] - knots[i - 1])
return float("nan")
def at(knots, f, p):
i = max(0, min(bisect.bisect_right(knots, p) - 1, len(knots) - 2))
t = (p - knots[i]) / (knots[i + 1] - knots[i])
return f[i] + t * (f[i + 1] - f[i])
def four_points(rows, convention="A"):
k, tc, ch, ex, te = curves(rows)
nch = [1 - v for v in ch]
nex = [1 - v for v in ex]
opp, ipp = cross(k, tc, te), cross(k, ch, ex)
if convention == "A":
pmc, pme = cross(k, tc, nch), cross(k, te, nex)
else:
pmc, pme = cross(k, tc, ex), cross(k, te, ch)
return pmc, opp, ipp, pme
if __name__ == "__main__":
rows = survey()
k, tc, ch, ex, te = curves(rows)
print(" respondents after cleaning {}; distinct answered prices {}".format(len(rows), len(k)))
print(" price too_cheap cheap not_cheap expensive not_exp too_exp")
for p in range(40, 341, 30):
i = bisect.bisect_right(k, p) - 1
print(" {:5d} {:10.3f} {:7.3f} {:10.3f} {:10.3f} {:8.3f} {:8.3f}".format(
p, tc[i], ch[i], 1 - ch[i], ex[i], 1 - ex[i], te[i]))
for conv in ("A", "B"):
pmc, opp, ipp, pme = four_points(rows, conv)
print(" convention {}: PMC {:6.1f} OPP {:6.1f} IPP {:6.1f} PME {:6.1f} range width {:5.1f}"
.format(conv, pmc, opp, ipp, pme, pme - pmc))
pmc, opp, ipp, pme = four_points(rows, "A")
print(" joined curves: at OPP too cheap = too expensive = {:.3f}; at IPP cheap = expensive = {:.3f}"
.format(at(k, tc, opp), at(k, ch, ipp)))
print(" at PMC too cheap = not cheap = {:.3f}; at PME too expensive = not expensive = {:.3f}".format(
at(k, tc, pmc), at(k, te, pme)))
lost = sum(1 for r in rows if not (r[0] < opp < r[3]))
print(" respondents for whom OPP is outside their own (too cheap, too expensive) band: {} of {}"
.format(lost, len(rows)))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《零历史数据冷启动测价、范氏价格敏感度边界与贝叶斯自适应调价母典》第 2 章「范氏价格敏感度:四个问题、四个交点与它们管不到的地方」(版本 0.1.0). 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-133 · v0.1.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026133,
author = {Fan, Yafu},
title = {Zero-History Cold-Start Price Testing, Van Westendorp Sensitivity Bounds and Bayesian Adaptive Repricing},
year = {2026},
publisher = {5D Library},
edition = {0.1.0},
note = {DOI pending}
}