关于「实证闭环:买手私域的分层触达、贵宾关怀日历与证伪方案」,《邓巴分层与私域半衰期》给出的定量结论是什么?

结论:第八章给出买手私域的分层流程、贵宾关怀日历、小工具「私域分层排程器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。

数据口径:书中的层级下限、半衰期、权重与停止概率均为示意参数;全部计算由书中程序生成;本书是非个人化经营建议,合规规则以各地现行法规为准。 书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。

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: 19

layer sizes of a 1500 list20 40 120 360
budget 2.0 h/week scheduled20 40 24 0
budget h/wk VIP core reg light sched respond-only used
0.57 0 0 0 7 1493 0.48
1.014 0 0 0 14 1486 0.96
2.020 40 24 0 84 1416 2.00
2.520 40 120 360 540 960 2.45
4.020 40 120 360 540 960 2.45
half-life 30 d, 2.0 h/week19 0 0 0
ratio 4, 2.0 h/week sizes20 60 240 960
ratio 4, 2.0 h/week sched20 44 0 0
查看完整验算程序
# Chapter 8. p82: the scheduler core shared with distill/tool.py
# Four cumulative layers inner * ratio ** k (20, 60, 180, 540 of a
# 1500 list), floors 0.8 / 0.6 / 0.4 / 0.2, half-life 60 days, beta
# 0.5, interval = max(T*, 7 days), minutes per touch 6 / 3 / 1 / 0.5
# (all illustrative, as in p32 and p61). The weekly budget is filled
# inner layer first and an outer layer gets nobody while an inner one
# is cut; whoever does not fit becomes respond-only. The
# same functions run in the tool; here the budget is swept.
import math

NAMES = ("VIP", "core", "regular", "light")
FLOORS = (0.8, 0.6, 0.4, 0.2)


def layer_sizes(total, inner, ratio):
    sizes, prev = [], 0
    for k in range(4):
        cum = min(total, int(round(inner * ratio ** k)))
        sizes.append(max(0, cum - prev))
        prev = max(prev, cum)
    return sizes


def t_star(floor, lam, beta):
    return math.log((beta + floor * (1.0 - beta)) / floor) / lam


def plan(total=1500, inner=20, ratio=3.0, half_life=60.0, beta=0.5,
         budget=2.0, cap_days=7.0, minutes=(6.0, 3.0, 1.0, 0.5)):
    lam = math.log(2.0) / half_life
    sizes = layer_sizes(total, inner, ratio)
    keeps, used = [], []
    for n, floor, mins in zip(sizes, FLOORS, minutes):
        every = max(t_star(floor, lam, beta), cap_days)
        per_person = 365.0 / every * mins / 60.0 / 52.0
        left = max(0.0, budget - math.fsum(used))
        keep = n if n * per_person <= left else int(left / per_person)
        if keeps and keeps[-1] < sizes[len(keeps) - 1]:
            keep = 0
        used.append(keep * per_person)
        keeps.append(keep)
    return sizes, keeps, math.fsum(used)


def main():
    sizes, keeps, used = plan()
    print("layer sizes of a 1500 list: " + " ".join(map(str, sizes)))
    print("budget 2.0 h/week scheduled: " + " ".join(map(str, keeps)))
    print("budget h/wk  VIP core  reg light  sched  respond-only  used")
    for b in (0.5, 1.0, 2.0, 2.5, 4.0):
        s, k, u = plan(budget=b)
        tot = k[0] + k[1] + k[2] + k[3]
        print("{:11.1f} {:4d} {:4d} {:4d} {:5d} {:6d} {:13d} {:5.2f}"
              .format(b, k[0], k[1], k[2], k[3], tot, 1500 - tot, u))
    s, k, u = plan(half_life=30.0, budget=2.0)
    print("half-life 30 d, 2.0 h/week : " + " ".join(map(str, k)))
    s, k, u = plan(ratio=4.0, budget=2.0)
    print("ratio 4, 2.0 h/week sizes  : " + " ".join(map(str, s)))
    print("ratio 4, 2.0 h/week sched  : " + " ".join(map(str, k)))


main()

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《邓巴数分层、关系维护时间预算与私域客户衰减半衰期母典》第 8 章「实证闭环:买手私域的分层触达、贵宾关怀日历与证伪方案」(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-231 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026231,
  author    = {Fan, Yafu},
  title     = {Dunbar's Layered Numbers, Relationship-Maintenance Time Budgets and the Decay Half-Life of Private-Channel Customers},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论