关于「缓冲放在链尾:关键链、肥尾模块与接单边界」,《演示先行与四十八小时交付》给出的定量结论是什么?
结论:逐项加垫让计划虚胖,又因为早完成不交棒而只吃坏消息:示意六模块订单的加垫计划 15.91 小时,准时概率只有 0.721。链尾缓冲按中位排、早完成即交棒,承诺 14.73 小时,准时概率 0.997。平方和开方法会低估两条线并行时的缓冲,只能当下限。目录外模块有肥尾,加一个计划 3 小时的定制接入就能把准时概率拖到 0.632,所以一律移出四十八小时订单。
数据口径:书中的赢率、工时、毛利、首付比例、月费与价格均为示意参数,以「单位」计,不对应任何币种、真实客户或真实成交;全部计算由书中程序生成。 本书不是法律意见,也不是个人化经营或理财建议,合同条款须由当地执业律师审阅;书中不写威胁催促、不写会重置的倒计时、不写收益承诺,推荐奖励只写单层。
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: 0.721, 0.997, 14.73, 15.91
| regime A, padded plan | 15.91 h |
| P(done by padded plan) | 0.645, P(on time 16 h) = 0.721 |
| mean with no early hand-on | 16.01 h |
| regime B, median plan | 12.09 h + buffer |
| simulated 95 % buffer 2.65 h -> commit 14.73 h | |
| root-sum-square buffer 1.99 h -> commit 14.08 h | |
| P(on time 16 h) | 0.997, mean 12.80 h |
| padding minus pooled commitment | 1.18 h |
查看完整验算程序
# Chapter 4. p41: pad every module, or pool one buffer at the chain end?
# Six-module order and lanes of p32/p33. Regime A pads each module to
# its own P90 and, as Parkinson's law and late starts suggest, gains
# from early finishes are not passed on: a module ends at
# max(actual, padded). Regime B (critical chain, Goldratt 1997) plans
# each module at its median, passes early finishes on, and keeps one
# project buffer at the end of the chain. Root-sum-square sizing is
# compared with the simulated 95 % buffer. Illustrative; N = 20,000.
import math
import random
N = 20000
WINDOW = 16.0
LANE_A = [(0.5, 1.0, 2.0), (3.0, 4.0, 7.0), (1.5, 2.5, 5.0)]
LANE_B = [(2.0, 3.0, 5.0), (2.0, 3.0, 6.0), (1.0, 2.0, 4.0)]
QA = (2.0, 3.0, 4.0)
def tri_q(m, u):
lo, mode, hi = m
cut = (mode - lo) / (hi - lo)
if u <= cut:
return lo + math.sqrt(u * (hi - lo) * (mode - lo))
return hi - math.sqrt((1 - u) * (hi - lo) * (hi - mode))
def lane_time(mods, draws, floor_q):
parts = []
for m, d in zip(mods, draws):
parts.append(max(d, tri_q(m, floor_q)) if floor_q else d)
return math.fsum(parts)
rng = random.Random(20261120)
pad_t, pool_t = [], []
for _ in range(N):
da = [rng.triangular(m[0], m[2], m[1]) for m in LANE_A]
db = [rng.triangular(m[0], m[2], m[1]) for m in LANE_B]
dq = rng.triangular(QA[0], QA[2], QA[1])
pad_t.append(max(lane_time(LANE_A, da, 0.9),
lane_time(LANE_B, db, 0.9)) + max(dq, tri_q(QA, 0.9)))
pool_t.append(max(math.fsum(da), math.fsum(db)) + dq)
pool_t.sort()
def plan_at(u):
a = math.fsum(tri_q(m, u) for m in LANE_A)
b = math.fsum(tri_q(m, u) for m in LANE_B)
return max(a, b) + tri_q(QA, u)
padded, median = plan_at(0.9), plan_at(0.5)
sim_buf = pool_t[int(0.95 * N)] - median
heavy = LANE_A if math.fsum(tri_q(m, 0.5) for m in LANE_A) >= \
math.fsum(tri_q(m, 0.5) for m in LANE_B) else LANE_B
rss = math.sqrt(math.fsum((tri_q(m, 0.9) - tri_q(m, 0.5)) ** 2
for m in heavy + [QA]))
print("regime A, padded plan: {:.2f} h".format(padded))
print(" P(done by padded plan) = {:.3f}, P(on time 16 h) = {:.3f}".format(
len([t for t in pad_t if t <= padded]) / N,
len([t for t in pad_t if t <= WINDOW]) / N))
print(" mean with no early hand-on: {:.2f} h".format(
math.fsum(pad_t) / N))
print("regime B, median plan: {:.2f} h + buffer".format(median))
print(" simulated 95 % buffer {:.2f} h -> commit {:.2f} h".format(
sim_buf, median + sim_buf))
print(" root-sum-square buffer {:.2f} h -> commit {:.2f} h".format(
rss, median + rss))
print(" P(on time 16 h) = {:.3f}, mean {:.2f} h".format(
len([t for t in pool_t if t <= WINDOW]) / N, math.fsum(pool_t) / N))
print("padding minus pooled commitment: {:.2f} h".format(
padded - (median + sim_buf)))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《半成品演示先行逼单、四十八小时交钥匙系统交付与防需求蔓延母典》第 4 章「缓冲放在链尾:关键链、肥尾模块与接单边界」(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-368 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026368,
author = {Fan, Yafu},
title = {Demo-First B2B Closing, Forty-Eight-Hour Turnkey System Delivery and Scope-Creep Containment},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}