5D Library › 分拆经济学、防套利身份锁定与多阶授权阶梯 · English

第五章 六级菜单:一个精确求解的甄别问题

从引流钩子到总代席位

作者提纲中的六级菜单是:引流钩子 p1p_1、单次深度报告 p2p_2、单板块个人年卡 p3p_3、跨板块正交加购 p4p_4、开放多客排盘并可提现 50% 佣金的商业主理人牌照 p5p_5、私董会与全矩阵总代席位 p6p_6。卖家不知道每位来访者属于哪一类,只能摆出一张菜单,让每一类人自己挑。这正是 Mussa 与 Rosen 的非线性定价(甄别)问题:卖家设计一组「品质—价格」组合,使每一类买家都愿意选给自己设计的那一项(激励相容,IC),并且愿意购买(个人理性,IR)(Mussa and Rosen 1978)。

六类买家

程序设六类来访者:好奇者、求解者、常客、深度爱好者、工作室与总代,占比依次为 0.60、0.22、0.10、0.05、0.025 与 0.005,各类对六级产品的估值(美元)列在程序中。工作室对牌照的估值来自它的佣金收入:每年 260 位客户、每位 99、佣金 0.5,合计 12,870。估值满足递增差分(单交叉条件):越高的类型,从低一级升到高一级时多得的价值越多。每一级有单位成本(牌照与席位含培训与支持)。

在递增差分之下,最优菜单的价格由一条链决定:最低被服务类型的 IR 约束取等号,此后每一类相对下一类的「向下 IC」约束取等号: pj=pj−1+vt(j)−vt(j−1).p_j = p_{j-1} + v_t(j) – v_t(j-1) . 程序枚举所有单调的「类型—层级」分配,按这条链定价,再逐一核对每一类的选择,取利润最高者,所以结果是精确最优解,而不是近似。

推导

# Chapter 5. [p09] The six-tier menu as a screening problem.
# Tiers: 1 hook, 2 deep report, 3 one-module annual card, 4 cross-module
# add-on (two modules), 5 business licence (multi-client charts, 50%
# commission withdrawal), 6 private board + full-matrix agency seat.
# Six buyer types, shares per visitor. Values in USD (willingness to
# pay for each tier). A studio's licence value is its commission income:
# 0.5 x 99 x 260 clients a year = 12,870. Values have increasing
# differences (single crossing). Unit costs per tier below.
# Exact search: every monotone assignment of types to tiers (or none),
# prices from the binding IR of the lowest type served and the binding
# downward IC of each next type, then every choice re-checked.
from itertools import combinations_with_replacement as cwr

TYPES = ("curious", "seeker", "regular", "enthusiast", "studio",
         "agency")
SHARE = (0.60, 0.22, 0.10, 0.05, 0.025, 0.005)
V = ((12, 30, 40, 45, 50, 55),
     (15, 120, 180, 210, 240, 260),
     (18, 160, 330, 420, 520, 560),
     (20, 180, 420, 700, 1100, 1300),
     (20, 200, 500, 900, 12870, 16000),
     (20, 200, 520, 1000, 16000, 38000))
COST = (1, 3, 8, 12, 1500, 6000)

def prices_for(assign):
    price, prev = {}, None
    for t, j in enumerate(assign):
        if j == 0:
            continue
        if prev is None:
            price[j] = V[t][j - 1]
        elif j not in price:
            price[j] = price[prev] + V[t][j - 1] - V[t][prev - 1]
        prev = j
    return price

def choose(t, price):
    """the tier with the largest surplus >= 0; ties go up the ladder"""
    opts = [(round(V[t][j - 1] - price[j], 9), j) for j in price]
    opts = [o for o in opts if o[0] >= 0]
    return max(opts)[1] if opts else 0

def profit(price):
    tot, by = 0.0, {}
    for t in range(6):
        j = choose(t, price)
        if j:
            m = SHARE[t] * (price[j] - COST[j - 1])
            tot += m
            by[j] = by.get(j, 0.0) + m
    return tot, by

def solve(allowed=(1, 2, 3, 4, 5, 6)):
    best = (-1, None, None)
    for assign in cwr((0,) + allowed, 6):
        price = prices_for(assign)
        p, by = profit(price)
        if p > best[0] + 1e-9:
            best = (p, price, by)
    return best

if __name__ == "__main__":
    p, price, by = solve()
    print("   optimal menu (profit {:.2f} per visitor)".format(p))
    print("   tier  price     ratio to previous  contribution share")
    prev = None
    for j in sorted(price):
        r = "" if prev is None else "{:.2f}".format(price[j] / prev)
        print("   {:4d} {:9.2f} {:>12s} {:20.3f}".format(
            j, price[j], r, by.get(j, 0.0) / p))
        prev = price[j]
    for t, name in enumerate(TYPES):
        j = choose(t, price)
        print("     {:10s} buys tier {}  surplus {:6.0f}".format(
            name, j, V[t][j - 1] - price[j] if j else 0.0))
#    optimal menu (profit 505.94 per visitor)
#    tier  price     ratio to previous  contribution share
#       1     12.00                             0.013
#       2    117.00         9.75                0.050
#       3    287.00         2.45                0.055
#       4    567.00         1.98                0.055
#       5  12537.00        22.11                0.545
#       6  34537.00         2.75                0.282
#      curious    buys tier 1  surplus      0
#      seeker     buys tier 2  surplus      3
#      regular    buys tier 3  surplus     43
#      enthusiast buys tier 4  surplus    133
#      studio     buys tier 5  surplus    333
#      agency     buys tier 6  surplus   3463
六级菜单的价格(右轴,对数刻度)与各级对利润的贡献。

最优菜单是:12、117、287、567、12537、34537,价格确实跨越三个数量级;商业牌照约为一万、总代席位约为三万,与提纲的设想一致。相邻台阶的价格比是 9.75、2.45、1.98、22.11 与 2.75。利润贡献却高度集中:牌照占 0.545,席位占 0.282,两级合计约八成;前四级合计不到两成。

两级 B2B 之间的价格差同样由 IC 链决定:总代席位的价格,是牌照价格加上总代从牌照升级到席位所多得的价值。工作室在牌照上保留 333 的剩余,总代在席位上保留 3463 的剩余:这就是甄别模型中「信息租」,高类型为了不冒充低类型而必须得到的让利。

菜单有多依赖顶端

程序在五种变化下重新精确求解。

推导

# Chapter 5. [p10] How much of the ladder's profit depends on its top?
# The model of p09 (same six types, values, shares and unit costs) is
# re-solved exactly under five changes: (a) consumer tiers only (1-4);
# (b) no agency seat (1-5); (c) licence capped at 3,000 because "it is
# only software" (it stays on the menu, higher prices follow from it);
# (d) studio and agency shares halved; (e) the studio commission
# halved (licence value 0.25 x 99 x 260 = 6,435).
from itertools import combinations_with_replacement as cwr

SHARE = (0.60, 0.22, 0.10, 0.05, 0.025, 0.005)
V = ((12, 30, 40, 45, 50, 55),
     (15, 120, 180, 210, 240, 260),
     (18, 160, 330, 420, 520, 560),
     (20, 180, 420, 700, 1100, 1300),
     (20, 200, 500, 900, 12870, 16000),
     (20, 200, 520, 1000, 16000, 38000))
COST = (1, 3, 8, 12, 1500, 6000)

def solve(v, share, allowed=(1, 2, 3, 4, 5, 6), cap=None):
    best = (-1.0, None)
    for assign in cwr((0,) + allowed, 6):
        if cap and not set(cap) <= set(assign):
            continue          # a capped tier must stay on the menu
        price, prev = {}, None
        for t, j in enumerate(assign):
            if j == 0:
                continue
            if prev is None:
                price[j] = v[t][j - 1]
            elif j not in price:
                price[j] = price[prev] + v[t][j - 1] - v[t][prev - 1]
            if cap and j in cap:
                price[j] = min(price[j], cap[j])
            prev = j
        tot = 0.0
        for t in range(6):
            opts = [(round(v[t][j - 1] - price[j], 9), j) for j in price]
            opts = [o for o in opts if o[0] >= 0]
            if opts:
                j = max(opts)[1]
                tot += share[t] * (price[j] - COST[j - 1])
        if tot > best[0] + 1e-9:
            best = (tot, price)
    return best

if __name__ == "__main__":
    base, _ = solve(V, SHARE)
    v_half = [list(r) for r in V]
    v_half[4][4] = 6435
    cases = (("p09 menu", V, SHARE, (1, 2, 3, 4, 5, 6), None),
             ("(a) tiers 1-4 only", V, SHARE, (1, 2, 3, 4), None),
             ("(b) no agency seat", V, SHARE, (1, 2, 3, 4, 5), None),
             ("(c) licence <= 3,000", V, SHARE, (1, 2, 3, 4, 5, 6),
              {5: 3000}),
             ("(d) B2B shares halved", V,
              SHARE[:4] + (0.0125, 0.0025), (1, 2, 3, 4, 5, 6), None),
             ("(e) commission halved", v_half, SHARE,
              (1, 2, 3, 4, 5, 6), None))
    print("   case                      profit/visitor  vs p09   p5     p6")
    for name, v, sh, al, cap in cases:
        p, price = solve(v, sh, al, cap)
        print("   {:24s} {:12.2f} {:8.3f} {:>6s} {:>6s}".format(
            name, p, p / base,
            "{:.0f}".format(price[5]) if 5 in price else "-",
            "{:.0f}".format(price[6]) if 6 in price else "-"))
    print("\nReading: the two B2B tiers carry most of the profit. A cheap")
    print("licence does not just lose its own margin: the agency price")
    print("is tied to it through the IC chain and falls with it. If the")
    print("commission halves, the best menu drops the licence and sells")
    print("studios the agency seat at a lower price instead.")
#    case                      profit/visitor  vs p09   p5     p6
#    p09 menu                       505.94    1.000  12537  34537
#    (a) tiers 1-4 only             103.98    0.206      -      -
#    (b) no agency seat             418.44    0.827  12537      -
#    (c) licence <= 3,000           219.83    0.434   3000  25000
#    (d) B2B shares halved          296.63    0.586  12537  34537
#    (e) commission halved          377.34    0.746      -  15667
#
# Reading: the two B2B tiers carry most of the profit. A cheap
# licence does not just lose its own margin: the agency price
# is tied to it through the IC chain and falls with it. If the
# commission halves, the best menu drops the licence and sells
# studios the agency seat at a lower price instead.

只卖消费者层级(1 至 4 级)时,利润只剩 0.206;没有总代席位时剩 0.827。最值得注意的是(c):若因为「只是一个软件」而把牌照压到 3,000 以下,利润只剩 0.434,而且总代席位的价格也被拉到 25000,因为 IC 链把它锚定在牌照价格上。(e)若工作室的佣金减半,最优菜单干脆取消牌照,以 15667 把总代席位同时卖给工作室与总代。

假说 · 尚未验证

作者的诠释(尚未验证): 作者的生意本质上是一家 B2B 牌照公司,消费者层级的作用主要是筛选与培育未来的工作室。什么会证明这一判断错:若真实数据中工作室与总代对牌照与席位的估值远低于本书的参数(例如工作室每年只有几十位客户),p10 中的(e)就成了现实,最优菜单不再包含牌照,作者应当把重心放回消费者层级。

小结

把六级菜单写成甄别问题后,可以精确求出最优价格:在本书参数下为 12、117、287、567、12537、34537,牌照与总代席位贡献约八成利润。正因为如此,菜单对顶端的设定十分敏感:牌照定得太低,会经由 IC 链把席位价格一起拉低;佣金收入一旦减半,牌照这一级就不值得存在。

参考文献

Mussa, Michael, and Sherwin Rosen. 1978. “Monopoly and Product Quality.” Journal of Economic Theory 18 (2): 301–17.

Fan, Y. (2026). Unbundling Economics, Anti-Arbitrage Identity Locking and Multi-Tier Licensing Ladders: Orthogonal Demand Partitioning, Resale Cannibalization Bounds, Affiliate Margin Preservation and Five-Figure B2B Upsell Equilibria [分拆经济学、防套利身份锁定与多阶授权阶梯] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090184 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).