5D Library › Enterprise Bottleneck Diagnostics, SKU Cannibalization Audits and Automation Return Bounds · 中文

Preface: Find the Bottleneck before Changing Anything

Note

Please read this first. This book studies the mathematics of three things: where a company’s bottleneck is, and why automation or staff added at a non-bottleneck do not raise output; how sub-line and budget products cannibalize the flagship’s margin, and when delisting SKUs raises profit; and how to design an outside consultant’s contract of “a fixed base fee plus a share of the incremental profit”. Team capacities, demand parameters, margins, noise and contract parameters are illustrative or synthetic models, meant to show structure, not data from any real company; all data are generated by the programs printed in the book. The book is not investment, legal or tax advice.

Volume 79 dealt with following up buyers who did not buy and with screening high-ticket clients. This book turns to the inside of the firm: when a consultant is brought in to “diagnose” a company, what should be measured first, what changed first, and how should the work be paid for? Each question has a popular answer: automate a non-bottleneck blindly and work in progress and communication friction blow up quadratically; delist the least efficient 30% of SKUs and profit rises; consultants should take a share of the incremental profit. This book writes them as runnable arithmetic and then asks when they hold.

What the book contains

Chapter 1 writes the four teams of content, acquisition, conversion and fulfilment as a Jackson queueing network, finds the bottleneck, and uses mean value analysis of a closed network to show the ceiling on throughput. Chapter 2 computes what happens to work pushed into the bottleneck, the communication channels in a team, and how a non-bottleneck can help the bottleneck by reducing variability. Chapter 3 uses a nested logit model to compute diversion ratios and the change in profit from delisting SKUs, and Chapter 4 writes down the margin dilution matrix and tests “cut 30%” over 400 random worlds. Chapter 5 solves for the share under double moral hazard and for audit rules that deter hiding the gain, and Chapter 6 compares linear shares, stepped shares and flat fees and computes how many months it takes to measure the gain. Chapter 7 reads the Guanzi’s “Cheng Ma” and “Qing Zhong II”, the market exchange in the History of Song’s treatise on food and money, and the Han Feizi’s “Ding Fa”. Chapter 8 finds the bottleneck from event logs and gives a 24-month contract ledger and a pre-registered switchback experiment.

Claims corrected on checking

  1. Automating a non-bottleneck does not raise the ceiling on output, but it can shorten lead times. Doubling the content team’s capacity leaves the maximum throughput at 24.0 jobs a day and cuts the lead time from 0.617 to 0.512 days (p01).
  2. Work in progress does not “blow up quadratically”. It rises because more work is pushed into the bottleneck, where it grows along the hyperbola ρ/(1−ρ)\rho/(1-\rho): 49.00 jobs at a load of 0.98, against 3.84 for a quadratic matched at 0.5. What grows quadratically is the number of communication channels, n(n−1)/2n(n-1)/2 (p03).
  3. The more pushed in, the fewer closed. Leads go cold in the queue: sales peak when 25.5 leads a day are released (11.06 sales) and fall to 5.22 when 29.5 are released (p03).
  4. A non-bottleneck can help, but mostly by reducing variability. Halving the variability of the acquisition hand-over cuts the conversion team’s wait by 22%; doubling acquisition’s speed cuts it by only 5% (p04).
  5. “Cut the bottom 30%” is not a law. In the baseline model profit rises 5.3%, mainly from demand diverted to higher-margin products; once “a fuller line draws more visits” is added, a 30% cut by naive contribution loses money in 218 of 400 random worlds (p06, p08).
  6. A higher share is not always better. The company’s own execution needs an incentive too: the best share is about 0.214, and at a share of 0.50 the joint surplus is below that of a fixed fee alone (p09).
  7. Two sources. “Does the balance have a number?” and “the balance has no number” are in the Guanzi’s “Qing Zhong II”, not “Qing Zhong I”; the Han Feizi’s “Ding Fa” reads “follow the name and demand the reality” (Chapter 7).

How to read the boxes

Marked boxes separate published evidence, runnable Python calculations (16 programs, standard library only, fixed random seeds, each actually run at build time with its real output printed below it), classical sources, notes and the author’s hypotheses; every hypothesis states what result would show it wrong. Programs and their output use ASCII characters only.

Disclosure

The author uses the works of 5D Library, including this book, as the methodological basis of his community courses, the 5D Ascension Cards (Dimensional Shift Card Decks) and the Yafu AI knowledge base, and runs a number of websites, paid online reading software, subscriptions, reports, courses, business licences and business advisory services whose diagnosis work, product-line planning and consulting contracts are exactly what this book studies, so he has a direct commercial interest in its subject. The book promotes no software, course, card deck or service, uses no data from the author’s businesses or their clients, and accepts no sponsorship. Its parameters are synthetic or illustrative. The experiment in Chapter 8 must follow a pre-registered protocol, and its results will be published whether positive or negative.

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).