关于「漏斗单位经济:首单回本线为什么不是止损线」,《收费站买量精算》给出的定量结论是什么?

结论:四十条素材里有 23 条真正赚钱,全部留下时的利润是 17,241 令吉。规则 A 误杀了 15 条好素材,又放过 2 条坏素材,只拿到 3,822,占可得利润的 0.222。规则 B 好得多,误杀 7 条,可它放过了 13 条坏素材,拿到 12,295,占 0.713。规则 C 一条不错,拿满 1.000。

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: 0.222, 0.713, 1.000, 12295, 13, 15

site-average funnel: u 0.085 refund 0.151 V 19.88
good creatives 23 of 40; best attainable profit 17,241
rulekill goodkeep badprofitshare
A first ROAS>=11523,8220.222
B site-average V71312,2950.713
C own back-fill0017,2411.000
查看完整验算程序
# Chapter 1. p13: a first-order ROAS stop rule misjudges creatives.
# 40 synthetic creatives for one assessment paywall (RM, illustrative).
# Each creative has a true CPA, upsell rate and refund rate. "Bait"
# creatives buy cheap icebreakers from people who rarely upsell and
# often refund; "fit" creatives cost more per buyer but bring readers
# who buy the deep report. Truth: keep a creative when V_i > CPA_i.
# Rule A keeps it when first-order ROAS = P1 / CPA >= 1. Rule B keeps it
# when CPA <= V computed with site-average funnel rates. Rule C uses the
# creative's own back-filled upsell and refund rates (week 4 onwards).
# Profit is counted on RM 1,000 of spend per creative.
import math
import random

P1, P2, CB, CBF, RATE, FIX, TAX, C1, C2, REP = \
    9.90, 129.0, 0.005, 15.0, 0.034, 0.50, 0.06, 0.40, 1.50, 6.00


def order_net(price, refund, cost):
    kept = price * (1 - refund - CB) * (1 - TAX)
    return kept - (price * RATE + FIX) - CB * CBF - cost


def value(u, rf, rep):
    return math.fsum([order_net(P1, rf, C1), u * order_net(P2, 0.04, C2),
                      rep])


rng = random.Random(20261120)
ads = []
for i in range(40):
    if i % 2 == 0:            # bait
        cpa = rng.uniform(5.0, 14.0)
        u = rng.uniform(0.00, 0.04)
        rf = rng.uniform(0.15, 0.35)
        rep = rng.uniform(0.5, 2.0)
    else:                     # fit
        cpa = rng.uniform(9.0, 40.0)
        u = rng.uniform(0.10, 0.22)
        rf = rng.uniform(0.02, 0.08)
        rep = rng.uniform(4.0, 9.0)
    ads.append((cpa, u, rf, rep))

avg_u = math.fsum(a[1] for a in ads) / len(ads)
avg_rf = math.fsum(a[2] for a in ads) / len(ads)
avg_rep = math.fsum(a[3] for a in ads) / len(ads)
v_avg = value(avg_u, avg_rf, avg_rep)

rules = {"A first ROAS>=1": lambda c, v: P1 / c >= 1.0,
         "B site-average V": lambda c, v: c <= v_avg,
         "C own back-fill": lambda c, v: c <= v}
print("site-average funnel: u {:.3f} refund {:.3f} V {:.2f}".format(
    avg_u, avg_rf, v_avg))
truth_profit = []
for cpa, u, rf, rep in ads:
    v = value(u, rf, rep)
    truth_profit.append(1000.0 * (v / cpa - 1.0))
best = math.fsum(p for p in truth_profit if p > 0)
print("good creatives {:d} of 40; best attainable profit {:,.0f}".format(
    len([p for p in truth_profit if p > 0]), best))
print("  rule               kill good  keep bad   profit  share")
for name in rules:
    kg = kb = 0
    prof = []
    for (cpa, u, rf, rep), p in zip(ads, truth_profit):
        keep = rules[name](cpa, value(u, rf, rep))
        if keep:
            prof.append(p)
        if p > 0 and not keep:
            kg += 1
        if p <= 0 and keep:
            kb += 1
    tot = math.fsum(prof)
    print("  {:<18} {:>9d} {:>9d} {:>8,.0f} {:>6.3f}".format(
        name, kg, kb, tot, tot / best))

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 1 章「漏斗单位经济:首单回本线为什么不是止损线」(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