关于「扩量控制律:饱和曲线、滞留效应与边际回报」,《收费站买量精算》给出的定量结论是什么?

结论:第六章用 Hill 饱和曲线与几何滞留说明平均 ROAS 掩盖边际 ROAS:从每天 2,000 令吉加两成,新增部分的 ROAS 只有 0.769,平均 ROAS 只从 1.110 降到 1.053;按边际回报控制预算的律优于两成口诀,「日入百单」只作工程目标反推口袋数与上新速度。

数据口径:按示意参数,tarotmalaysia 一个破冰买家的期望净值是破冰价的 2.68 倍,按首单回本线止损,在四十条合成素材上只拿到可得利润的 0.222。 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: 0.769, 1.053, 1.110, 2000

A. steady state, theta 0.5, h 1.2
daily RM buyers avg ROAS marg ROAS profit RM
25010.6 1.126 1.286 31
50023.0 1.217 1.308 108
1,00046.4 1.231 1.166 231
1,50066.7 1.179 0.986 269
2,00083.8 1.110 0.825 220
3,000110.0 0.972 0.583 -84
4,000128.8 0.854 0.425 -585
6,000153.3 0.677 0.246 -1,937
B. marginal ROAS1 at RM 1,500 a day (profit peak)
average ROAS1 at RM 2,788 a day (profit zero)
C. a +20 % step from RM 2,000: increment ROAS 0.769
while average ROAS moves 1.1101.053
D. adstock: days to reach 90 % of a new spend level3.32
查看完整验算程序
# Chapter 6. p61: average ROAS hides marginal ROAS on a saturating curve.
# Daily buyers b(x) = BMAX e^h / (K^h + e^h), where e is the geometric
# adstock of spend (carry-over theta per day; in steady state at a
# constant daily spend x, e = x / (1 - theta)). Hill shape as in the
# media mix model of Jin et al. (2017). Value per buyer V = 26.51 RM
# (net, p11), so ROAS here is net-value ROAS. Illustrative parameters.
import math

BMAX, K, H, THETA, V = 220.0, 6000.0, 1.2, 0.5, 26.51


def buyers(x):
    e = x / (1.0 - THETA)
    return BMAX * e ** H / (K ** H + e ** H)


def marginal(x, d=1.0):
    return V * (buyers(x + d) - buyers(x - d)) / (2 * d)


def solve(f, lo, hi):
    for _ in range(200):
        mid = 0.5 * (lo + hi)
        if f(mid) > 0:
            lo = mid
        else:
            hi = mid
    return 0.5 * (lo + hi)


if __name__ == "__main__":
    print("A. steady state, theta {:.1f}, h {:.1f}".format(THETA, H))
    print("  daily RM  buyers  avg ROAS  marg ROAS  profit RM")
    for x in (250, 500, 1000, 1500, 2000, 3000, 4000, 6000):
        b = buyers(x)
        print("  {:>8,d} {:>7.1f} {:>9.3f} {:>10.3f} {:>10,.0f}".format(
            x, b, b * V / x, marginal(x), b * V - x))
    xm = solve(lambda x: marginal(x) - 1.0, 1500.0, 20000.0)
    xa = solve(lambda x: buyers(x) * V / x - 1.0, 1000.0, 50000.0)
    print("B. marginal ROAS = 1 at RM {:,.0f} a day (profit peak)"
          .format(xm))
    print("   average ROAS = 1 at RM {:,.0f} a day (profit zero)"
          .format(xa))
    print("C. a +20 % step from RM 2,000: increment ROAS {:.3f}".format(
        V * (buyers(2400) - buyers(2000)) / 400))
    print("   while average ROAS moves {:.3f} -> {:.3f}".format(
        buyers(2000) * V / 2000, buyers(2400) * V / 2400))
    print("D. adstock: days to reach 90 % of a new spend level: {:.2f}"
          .format(math.log(0.1) / math.log(THETA)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《测评与命理收费站买量精算、服务端转化回传校准与日入百单扩量母典》第 6 章「扩量控制律:饱和曲线、滞留效应与边际回报」(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-345 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026345,
  author    = {Fan, Yafu},
  title     = {Paid-Ad Scaling for Diagnostic Paywalls: Meta and TikTok Stopping Rules, Server-Side CAPI Calibration and Hundred-Order Daily Unit Economics},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论