5D Library › Affiliate Network Power Laws, Influencer Portfolio Optimization and Incentive Stability Bounds · 中文
Chapter 6. Cohort Decay and the Wholesale Margin of Licensed Principals
How a cohort drains away
Activation is only the start. p11 follows a thousand newly active affiliates for twenty-four months: weaker sellers stop sooner.
Calculation
# Chapter 6. p11: How an affiliate cohort decays. 1,000 newly active
# affiliates sell 10 x Lomax(1.5) units a month each (p02). Weak sellers
# quit sooner: the monthly chance of stopping is 0.05 + 0.25 / (1 + s/10)
# for an affiliate selling s a month. Tracked: the share still active,
# revenue relative to month 1 (net revenue retention), and the share of
# that month's revenue from the cohort's top 1% of sellers. 200 cohorts,
# seed 87.
import random
def cohort(rng, n=1000):
return [10.0 * ((1 - rng.random()) ** (-1 / 1.5) - 1) for _ in range(n)]
if __name__ == "__main__":
rng = random.Random(87)
marks = (1, 3, 6, 12, 24)
acc = {m: [0.0, 0.0, 0.0] for m in marks}
for _ in range(200):
s = cohort(rng)
top = set(sorted(range(1000), key=lambda i: -s[i])[:10])
alive = [True] * 1000
first = sum(s)
for month in range(1, 25):
if month in acc:
rev = sum(x for x, a in zip(s, alive) if a)
tr = sum(s[i] for i in top if alive[i])
acc[month][0] += sum(alive) / 1000
acc[month][1] += rev / first
acc[month][2] += tr / rev
for i in range(1000):
if alive[i] and rng.random() < 0.05 + 0.25 / (1 + s[i] / 10):
alive[i] = False
print(" month still active revenue vs month 1 top 1% share")
for m in marks:
a, r, t = (x / 200 for x in acc[m])
print(" {:5d} {:12.3f} {:18.3f} {:12.3f}".format(m, a, r, t))
# month still active revenue vs month 1 top 1% share
# 1 1.000 1.000 0.274
# 3 0.645 0.807 0.301
# 6 0.350 0.600 0.338
# 12 0.122 0.354 0.401
# 24 0.025 0.140 0.457After six months only 0.350 are still selling, yet revenue is still 0.600 of month one; after twelve months 0.122 of the people remain with 0.354 of the revenue; after twenty-four months only 0.025 remain, and revenue is still 0.140. The reason is that those who stay are mostly the strong: the initial top 1% of sellers bring 0.274 of revenue in month one and 0.457 by month twenty-four. So a programme’s long-run revenue depends mainly on keeping the top few, not on the number activated.
Where each hundred ringgit goes
p12 compares how a 100-ringgit retail sale is divided in three channels: the brand advertising directly, a single-tier affiliate on 50% commission, and a licensed principal who buys at 45 ringgit wholesale and sells retail herself.
Calculation
# Chapter 6. p12: Where each ringgit of a RM 100 sale goes, and a test
# for the legal line. Cost of goods RM 20. Direct: the brand pays
# RM 30 of ads per sale and RM 10 to ship. Single-tier affiliate: the
# brand pays 50% commission and RM 10 to ship. Licensee: buys at RM 45
# wholesale (the brand ships in bulk, RM 5 a unit) and spends RM 15 a
# unit selling to her own customers. Then a variant that crosses the
# line: each licensee also signs up two sub-licensees a month at a
# RM 5,000 fee (RM 2,000 to her) and takes 5% of their wholesale buying
# (200 units a month each). The test: what share of her income comes
# from recruiting rather than from sales to end customers?
P, COGS = 100.0, 20.0
if __name__ == "__main__":
rows = (("brand direct", P - COGS - 30 - 10, 0.0),
("single-tier affiliate", P * 0.5 - COGS - 10, P * 0.5),
("licensee wholesale", 45.0 - COGS - 5, P - 45 - 15))
print(" channel brand keeps partner keeps (RM)")
for name, brand, partner in rows:
print(" {:<22} {:11.0f} {:18.0f}".format(name, brand, partner))
sales = 300 * (P - 45 - 15)
print(" licensee selling 300 units a month earns RM {:,.0f} from"
" sales".format(sales))
print(" variant with paid sign-ups")
print(" month sub-licensees from recruiting share of income")
subs = 0
for month in range(1, 13):
subs += 2
rec = 2 * 2000.0 + 0.05 * 45 * 200 * subs
if month in (1, 3, 6, 12):
print(" {:5d} {:13d} {:15,.0f} {:15.3f}".format(
month, subs, rec, rec / (rec + sales)))
# channel brand keeps partner keeps (RM)
# brand direct 40 0
# single-tier affiliate 20 50
# licensee wholesale 20 40
# licensee selling 300 units a month earns RM 12,000 from sales
# variant with paid sign-ups
# month sub-licensees from recruiting share of income
# 1 2 4,900 0.290
# 3 6 6,700 0.358
# 6 12 9,400 0.439
# 12 24 14,800 0.552Selling direct, the brand keeps 40 ringgit a sale but bears the cost of finding customers itself; through a single-tier affiliate or a licensed principal the brand keeps 20, and the partner keeps 50 or 40. A licensed principal selling 300 units a month earns 12,000 ringgit from sales. All three arrangements are compliant, because everyone’s income comes from goods sold to end customers.
The trouble is the variant: if the licensed principal also signs up two “sub-principals” a month, each paying a 5,000-ringgit fee of which 2,000 goes to her, and takes 5% of their wholesale buying, then the share of her income that comes from recruiting is 0.290 in month one, 0.439 in month six and 0.552 in month twelve; past one half, it is already “primarily from recruitment”. The line is not crossed in one step but slowly, as the network grows; that is why the share has to be computed every month.
Note
A compliance monitoring measure: each month, compute for every licensed principal the “share of income from recruiting”, including any cut of sign-up fees, any percentage of downline wholesale buying and any bonus tied to recruitment; if this share is above zero at all, the plan contains pay tied to recruitment and that pay should be removed, not watched until it nears one half.
Summary
A cohort’s headcount decays fast and its revenue more slowly, because those who stay are the strong. Single-tier commission and the licensed principal’s wholesale margin both tie income to retail; once a cut of sign-up fees or a percentage of downline buying is added, the share of income from recruiting rises as the network grows and eventually crosses the legal line.
Fan, Y. (2026). Affiliate Network Power Laws, Influencer Portfolio Optimization and Incentive Stability Bounds: Heavy-Tailed Productivity, Markowitz Traffic Allocation and Single-Tier Commission Equilibria [分销网络幂律激活、网红矩阵组合优化与激励稳定边界] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090229 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).