关于「留存是生存曲线:Weibull 负老化、开场筛选与留存者分母」,《长视频半衰期与深度信任》给出的定量结论是什么?

结论:第四章把留存写成 Weibull 生存曲线,如实预告式开场在结尾留下的高意图观众是大承诺式的 1.125 倍。

数据口径:示意曲线下,短视频头 48 小时拿到三年观看的 0.750,带搜索地板的长视频只拿到 0.059,第 1,095 天仍有每天 5.00 次观看;半衰期翻十倍时,短内容峰值要高出 9.96 倍才打平现值。 书中参数为示意,全部计算由书中程序生成。

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: 1.125

per 10,000 impressionsA broadB honest
starters600450
30 s retention0.7600.730
completion rate0.1870.271
high-intent at 6 min136.7153.8
high-intent at 12 min108.3121.8
high-intent share at end0.9660.999
ratio B/A of high-intent viewers at 12 min: 1.125
查看完整验算程序
# Chapter 4. p42: the opening filters who stays; count the stayers.
# Two openings for the same 12-minute video (illustrative parameters):
#   A broad promise  : click rate 0.060; starters are 40 % high-intent
#   B honest preview : click rate 0.045; starters are 60 % high-intent
# (B states in the first seconds what will be computed and where the
# method fails.) High-intent viewers follow Weibull k = 0.5 with
# S(30 s) = 0.85 under both openings. Low-intent viewers follow
# k = 0.8 with S(30 s) = 0.70 under A (the broad promise holds them a
# little longer) and 0.55 under B. Counts are per 10,000 impressions.
import math

VIDEO = 720.0


def lam_for(k, s30):
    return 30.0 / (-math.log(s30)) ** (1.0 / k)


def surv(t, k, s30):
    return math.exp(-(t / lam_for(k, s30)) ** k)


OPEN = {"A": (0.060, 0.40, 0.70), "B": (0.045, 0.60, 0.55)}

if __name__ == "__main__":
    print("per 10,000 impressions   A broad   B honest")
    rows = {}
    for key, (ctr, hi, s30_lo) in OPEN.items():
        starts = 10000.0 * ctr
        nh, nl = starts * hi, starts * (1.0 - hi)

        def alive(t, nh=nh, nl=nl, s30_lo=s30_lo):
            return nh * surv(t, 0.5, 0.85), nl * surv(t, 0.8, s30_lo)

        h30, l30 = alive(30.0)
        h720, l720 = alive(VIDEO)
        h360, _ = alive(360.0)
        rows[key] = [starts, (h30 + l30) / starts, (h720 + l720) / starts,
                     h360, h720, h720 / (h720 + l720)]
    labels = ["starters", "30 s retention", "completion rate",
              "high-intent at 6 min", "high-intent at 12 min",
              "high-intent share at end"]
    fmt = ["{:>8,.0f}", "{:>8.3f}", "{:>8.3f}", "{:>8.1f}", "{:>8.1f}",
           "{:>8.3f}"]
    for i, lab in enumerate(labels):
        print("  {:<22} {}   {}".format(lab, fmt[i].format(rows["A"][i]),
                                       fmt[i].format(rows["B"][i])))
    print("ratio B/A of high-intent viewers at 12 min: {:.3f}".format(
        rows["B"][4] / rows["A"][4]))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《长视频搜索半衰期红利、深度信任预售与多年期被动获客资产母典》第 4 章「留存是生存曲线:Weibull 负老化、开场筛选与留存者分母」(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-370 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026370,
  author    = {Fan, Yafu},
  title     = {Long-Form Video Search Half-Lives, Deep-Trust Pre-Selling and Multi-Year Passive Acquisition Assets},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论