5D Library › Enterprise Bottleneck Diagnostics, SKU Cannibalization Audits and Automation Return Bounds · 中文
Chapter 2. Work Pushed into the Bottleneck, Communication Channels and Variability
The “quadratic blow-up” needs correcting
A common claim is that automating or adding staff at a non-bottleneck makes work in progress and communication friction “blow up quadratically”. On checking, two things must be separated. First, speeding up a non-bottleneck does not by itself add work in progress (in Chapter 1 it reduced it); work in progress rises because a faster upstream pushes more work into the bottleneck. Second, the work in progress of the bottleneck it is pushed into grows not quadratically but hyperbolically: a single-server queue holds on average, which goes to infinity as the load approaches 1. What does grow quadratically is the number of communication channels within a team, (Brooks 1975).
Calculation
# Chapter 2. p03: What pushing more work into the bottleneck does.
# (a) M/M/1 work in progress L = rho / (1 - rho) is hyperbolic in the
# load, not quadratic: it diverges as rho -> 1.
# (b) Faster upstream teams release more leads into conversion (3
# closers, 10 leads/day each). A lead waiting w days still closes with
# probability p0 * exp(-w / T): leads go cold. For M/M/c the queue
# wait has E[exp(-s W)] = (1 - Pw) + Pw * (c mu - lam) / (c mu - lam + s).
# (c) Coordination channels in a team of n people: n (n - 1) / 2.
from math import factorial
def erlang_c(lam, c, mu):
a, rho = lam / mu, lam / (c * mu)
s = sum(a ** k / factorial(k) for k in range(c))
top = a ** c / (factorial(c) * (1 - rho))
return top / (s + top)
C, MU, P0, T = 3, 10.0, 0.5, 1.0
def push(lam):
pw = erlang_c(lam, C, MU)
gap = C * MU - lam
wq = pw / gap # mean wait, days
lq = lam * wq
lt = (1 - pw) + pw * gap / (gap + 1 / T) # E[exp(-W/T)]
return lq + lam / MU, 24 * wq, lam * P0 * lt
if __name__ == "__main__":
print(" (a) rho L=rho/(1-rho) 0.5-matched quadratic 4*rho^2")
for r in (0.5, 0.8, 0.9, 0.95, 0.98):
print(" {:.2f} {:13.2f} {:10.2f}".format(
r, r / (1 - r), 4 * r * r))
print(" (b) leads/day WIP wait (h) closed sales/day")
best = max((push(l)[2], l) for l in [x / 2 for x in range(40, 60)])
for lam in (20.0, 24.0, 26.0, 28.0, 29.0, 29.5):
L, w, sold = push(lam)
print(" {:9.1f} {:5.1f} {:7.1f} {:8.2f}".format(
lam, L, w, sold))
print(" most sales at {:.1f} leads/day: {:.2f} per day".format(
best[1], best[0]))
print(" (c) team size channels")
for n in (4, 6, 8, 12):
print(" {:9d} {:8d}".format(n, n * (n - 1) // 2))
# (a) rho L=rho/(1-rho) 0.5-matched quadratic 4*rho^2
# 0.50 1.00 1.00
# 0.80 4.00 2.56
# 0.90 9.00 3.24
# 0.95 19.00 3.61
# 0.98 49.00 3.84
# (b) leads/day WIP wait (h) closed sales/day
# 20.0 2.9 1.1 9.60
# 24.0 5.0 2.6 10.89
# 26.0 7.5 4.6 11.03
# 28.0 15.1 10.5 9.91
# 29.0 30.1 22.5 7.70
# 29.5 60.1 46.5 5.22
# most sales at 25.5 leads/day: 11.06 per day
# (c) team size channels
# 4 6
# 6 15
# 8 28
# 12 66Compare the hyperbola with a quadratic matched to it at : at a load of 0.80 work in progress is 4.00 jobs against the quadratic’s 2.56; at 0.90 it is 9.00 against 3.24; at 0.95, 19.00 against 3.61; at 0.98, 49.00 against 3.84. The closer to saturation, the more dangerous the hyperbola is than the quadratic.
The more pushed in, the fewer closed
The conversion team has three closers, each handling 10 leads a day. As the upstream gets faster, more leads a day are released into it. A lead goes cold while it waits: after waiting days it closes with probability . Releasing 20.0 leads a day, the queue holds 2.9 leads, the mean wait is 1.1 hours and 9.60 sales close a day; releasing 26.0, 11.03 close; releasing 28.0, work in progress is 15.1 leads, the wait 10.5 hours, and sales fall to 9.91; releasing 29.5, work in progress is 60.1 leads, the wait 46.5 hours, and only 5.22 sales close. The release rate with the most sales is 25.5 leads a day (11.06 sales), well below the team’s nominal capacity.
Adding people to a non-bottleneck team has a second cost: a team of 8 has 28 pairwise channels of communication, and at 12 people it has 66. Half again as many people, more than twice as many channels.
Note
Release at the pace of the bottleneck. The theory of constraints calls this “drum, buffer, rope”: the pace at which the upstream releases work is set by the bottleneck, not by the upstream’s own capacity (Goldratt and Cox 1984). Here that means the advertising budget should be set by the number of leads the conversion team can absorb each day, not by the number the acquisition team can produce.
A non-bottleneck can help: by reducing variability
Improving a non-bottleneck does have a legitimate value, but usually not from speed: it comes from making the work that reaches the bottleneck smoother. Kingman’s formula relates queueing delay to variability (Kingman 1961), and the linking equation gives the variability of a station’s departures as a blend of its arrival and service variability (Hopp and Spearman 2008).
Calculation
# Chapter 2. p04: A non-bottleneck CAN help - by smoothing, not speed.
# Acquisition (one server) feeds conversion (3 closers, the bottleneck).
# Kingman/Sakasegawa queue wait for G/G/c:
# Wq = (ca2 + cs2)/2 * rho^(sqrt(2(c+1)) - 1) / (c (1 - rho)) * te
# and the linking equation for the variability of departures
# cd2 = 1 + (1 - rho^2)(ca2 - 1) + rho^2 (cs2 - 1) / sqrt(c).
# Leads arrive at 25 per day with ca2 = 1.5 (bursty campaigns).
from math import sqrt
LAM, CA2 = 25.0, 1.5
def station(lam, ca2, c, mu, cs2):
rho = lam / (c * mu)
te = 1 / mu
wq = (ca2 + cs2) / 2 * rho ** (sqrt(2 * (c + 1)) - 1) / (
c * (1 - rho)) * te
cd2 = 1 + (1 - rho ** 2) * (ca2 - 1) + rho ** 2 * (cs2 - 1) / sqrt(c)
return wq, cd2, rho
def line(mu_acq, cs2_acq, cs2_conv=1.0):
w1, cd2, r1 = station(LAM, CA2, 1, mu_acq, cs2_acq)
w2, _, r2 = station(LAM, cd2, 3, 10.0, cs2_conv)
return 24 * w1, cd2, 24 * w2
if __name__ == "__main__":
cases = (("base", 40.0, 2.0),
("acquisition twice as fast", 80.0, 2.0),
("acquisition half as variable", 40.0, 0.5))
b = line(40.0, 2.0)
print(" case acq wait cd2 conv wait (h)")
for name, mu, cs2 in cases:
w1, cd2, w2 = line(mu, cs2)
print(" {:<29} {:7.2f} {:5.3f} {:7.2f} ({:+.0%})".format(
name, w1, cd2, w2, w2 / b[2] - 1))
# case acq wait cd2 conv wait (h)
# base 1.75 1.695 4.63 (+0%)
# acquisition twice as fast 0.24 1.549 4.38 (-5%)
# acquisition half as variable 1.00 1.109 3.63 (-22%)The acquisition team launches advertising in batches and hands over leads in uneven bursts. In the original plan the conversion team’s mean wait is 4.63 hours. Doubling the acquisition team’s speed cuts its own wait from 1.75 hours to 0.24, but the conversion wait only to 4.38 hours (-5%). Without any speed-up, halving the variability of the hand-over (say fixed hand-over times and equal batch sizes) cuts the squared coefficient of variation of the departures from 1.695 to 1.109 and the conversion wait to 3.63 hours (-22%).
Hypothesis · not established
The author’s hypothesis (not yet tested): in small content-led acquisition teams, changing the hand-over from acquisition to conversion to fixed times and fixed batch sizes shortens the first response time to a lead more than doubling the acquisition team’s capacity. What would show it wrong: in an experiment switching at random week by week, the first response time under the former is not shorter than under the latter.
Summary
Speeding up a non-bottleneck adds no work in progress by itself; work in progress swells because more work is pushed into the bottleneck, and it grows along the hyperbola , more dangerously than a quadratic. What grows quadratically is the number of communication channels. Pushing in too much lets leads go cold and sales fall. A non-bottleneck’s greatest help to the bottleneck is often to make the work it hands over smoother.
References
Fan, Y. (2026). Enterprise Bottleneck Diagnostics, SKU Cannibalization Audits and Automation Return Bounds: Theory-of-Constraints Queueing Networks, Margin Dilution Matrices and Value-Based Advisory Pricing [企业瓶颈诊断、产品线自蚕食审计与自动化改造回报边界] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090217 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).