关于「零投流测试矩阵:十二条普拉克特—伯曼设计与样本量」,《短视频流量池冷启动》给出的定量结论是什么?

结论:第六章给出每周十二条、三个因子的零投流普拉克特—伯曼矩阵,主效应方差为单条的 0.333 倍,测出完播率 0.04 的差别需要 11 周。

数据口径:书中所有门槛、池子规模与权重均为示意,平台规则以其官方公示为准;全部计算由书中程序生成;本书是非个人化经营建议,只讨论真实内容质量与合规测试,不涉及任何刷量或规避平台规则的做法。 第三章说明硬门槛在数据上是斜坡,批次从 100 人增至 3,000 人时放行区间宽度从 0.124 收窄到 0.023;示意基准视频的破圈概率为 0.03585,瓶颈在第四道门。

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。 与结论吻合的打印值: 0.333, 11

video A hook B length C topic
1 result40s how-to
2 question40s story
3 result15s story
4 question40s how-to
5 question15s story
6 question15s how-to
7 result15s how-to
8 result40s how-to
9 result40s story
10 question40s story
11 result15s story
12 question15s how-to
factor A: +1 in 6 videos, -1 in 6
factor B: +1 in 6 videos, -1 in 6
factor C: +1 in 6 videos, -1 in 6
columns A.B dot product0
columns A.C dot product0
columns B.C dot product0
interaction alias: AB on C -0.333, AC on B -0.333, BC on A -0.333
variance of a main effect4/12 sigma^2 = 0.333 sigma^2
one factor at a time, 3 baseline + 3 per factor:
variance1/3 + 1/3 = 0.667 sigma^2, twice the matrix
查看完整验算程序
# Chapter 6. p61: the twelve-run Plackett-Burman matrix (1946).
# Twelve videos a week, three two-level factors (illustrative):
#   A hook   : question opening (-1) or result-first opening (+1)
#   B length : about 15 s (-1) or about 40 s (+1)
#   C topic  : how-to (-1) or behind-the-scenes story (+1)
# Rows are the cyclic shifts of the Plackett-Burman generator plus a
# row of all -1; we keep the first three columns. Every factor is +1
# in six videos and -1 in six, and every pair of columns is
# orthogonal, so each main effect is estimated from all twelve videos
# at once: effect = mean(+1) - mean(-1), variance 4 sigma^2 / 12.
# The price: each two-factor interaction leaks into the third factor's
# main effect with weight -1/3 (partial aliasing, printed below).
GEN = [1, 1, -1, 1, 1, 1, -1, -1, -1, 1, -1]
rows = [GEN[-i:] + GEN[:-i] for i in range(11)] + [[-1] * 11]
X = [r[:3] for r in rows]
label = {("A", -1): "question", ("A", 1): "result",
         ("B", -1): "15s", ("B", 1): "40s",
         ("C", -1): "how-to", ("C", 1): "story"}
print("video   A hook    B length  C topic")
for i, r in enumerate(X, 1):
    print("%5d   %-8s  %-8s  %s" % (i, label[("A", r[0])],
                                   label[("B", r[1])],
                                   label[("C", r[2])]))
cols = list(zip(*X))
for j, name in enumerate("ABC"):
    print("factor %s: +1 in %d videos, -1 in %d"
          % (name, cols[j].count(1), cols[j].count(-1)))
for a in range(3):
    for b in range(a + 1, 3):
        dot = 0
        for u, v in zip(cols[a], cols[b]):
            dot += u * v  # step: integer count
        print("columns %s.%s dot product = %d" % ("ABC"[a], "ABC"[b], dot))
alias = []
for a, b, o in ((0, 1, 2), (0, 2, 1), (1, 2, 0)):
    tot = 0
    for i in range(12):
        tot += cols[a][i] * cols[b][i] * cols[o][i]  # step: integer
    alias.append("%s%s on %s %+.3f" % ("ABC"[a], "ABC"[b], "ABC"[o],
                                        tot / 12.0))
print("interaction alias: " + ", ".join(alias))
print("variance of a main effect = 4/12 sigma^2 = 0.333 sigma^2")
print("one factor at a time, 3 baseline + 3 per factor:")
print("  variance = 1/3 + 1/3 = 0.667 sigma^2, twice the matrix")

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《全域短视频赛马池冷启动动力学、完播互动复播阶跃阈值与零投流破圈测试矩阵母典》第 6 章「零投流测试矩阵:十二条普拉克特—伯曼设计与样本量」(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-273 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026273,
  author    = {Fan, Yafu},
  title     = {Cold-Start Dynamics of Short-Video Traffic Pools: Step Thresholds in Completion, Engagement and Replay and a Zero-Ad-Spend Breakout Test Matrix},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论