06-reference

commoncog limits of operational excellence

Tue Apr 14 2026 20:00:00 GMT-0400 (Eastern Daylight Time) ·reference ·source: Commoncog ·by Cedric Chin

“The Limits of Operational Excellence” — @CedricChin

Why this is in the vault

This is the load-bearing caveat on Chin’s operational-excellence thesis — the answer to when Deming/SPC actually wins the business, and when it’s beside the point. The Danaher arc (Rales brothers, George Koenigsaecker, Larry Culp, Michael Milken’s junk bonds) makes the case that operational excellence is one leg of the Business Expertise Triad; Capital Allocation and Demand are the other two, and the operator who ignores them runs an excellent process to nowhere. Directly relevant to how RDCO should position the MAC discipline.

The core argument (paraphrased)

Operational excellence matters most when your business earns a return on capital near its cost of capital — not when it’s dying, and not when it’s a juggernaut.

Chin opens with the puzzle he left unresolved in Making Sense of Deming: why do some Deming-influenced companies acquire Process Power and compound, while others execute the same discipline and die? He reframes it three ways — “limits of Lean,” “limits of Deming,” and the most interesting: “when is operational excellence important for business success … and when is it irrelevant?”

The case is Danaher — S&P 100, 4,500% stock return over 25 years, widely considered one of the best operators in the world. The arc:

  1. Lean arrives by accident (1988). Steven and Mitchell Rales had built Danaher as a debt-fueled rollup of cheap manufacturing assets. Jake Brake GM George Koenigsaecker — frustrated with the division — met Toyota Production System engineers Yoshiki Iwata and Chihiro Nakao after a Hartford lecture and brought them to the factory that night. Lean worked, and the Rales forced it across the portfolio. The Danaher Production System became the Danaher Business System. Mark DeLuzio’s caveat: “you can’t copy Toyota” — DBS is constantly evolving and can’t be transplanted.
  2. Lean was paired with capital discipline from day one. The Rales and CFO Pat Allender were already good capital allocators. Lean freed cash flow; freed cash flow pushed ROIC up; ROIC was “nearly perfectly correlated” with cash generation. When risk-averse US banks wouldn’t fund acquisitions, the Rales went to Michael Milken and used high-yield “junk” bonds to buy 12 B2B companies between 1984 and 1990 — solid businesses, poorly run, cheap. The growth credited to Lean was actually Lean plus a debt-capital weapon competitors didn’t have.
  3. Culp’s refinement (2001) — the return sweet spot. CEO Larry Culp observed two things. First, “the efficacy of Lean was the highest … at a steady growth rate” — Lean yields 2–3% “free” capacity, doubled to 4–6% with modest capital; push beyond that and supply chains break. (Same number Jack Welch’s GE found with Six Sigma.) Second, highly cyclical businesses wasted Lean gains in recessions, so Culp pivoted acquisitions toward high-aftermarket, high-gross-margin businesses — Videojet, Radiometer, Beckman Coulter, ChemTreat. DBS worked better there because the margin structure left more fat to trim.
  4. The one-line answer (Brian Lui, on Twitter). “Operational excellence matters when the marginal/median company is earning near its cost of capital.” If you return $0.90–$1.01 on the dollar, operational excellence just helps you bleed out slower. If you’re Google earning $1.31 no matter how sloppy you are, it doesn’t meaningfully matter. The sweet spot is $1.03–$1.06 — where every percentage point of ROIC you wring out is real compounding value.

The deep implication (unstated but structural): this piece is cross-filed under Operations AND Capital because it’s arguing the two disciplines are inseparable. The Business Expertise Triad (Operations / Capital / Demand) is the frame. Danaher won because the Rales were excellent capital allocators who adopted Lean — not Lean operators who happened to raise debt. Most companies that “go Deming and die” have the operations leg and nothing else.

Mapping against Ray Data Co

This article is the load-bearing caveat for every RDCO deliverable. The Business Expertise Triad (Operations + Capital + Demand) is the framing RDCO should explicitly adopt — MAC and the harness are the operations leg, and the operations leg alone doesn’t build a durable business.

1. MAC is the operations leg of the triad — not the moat. The ../01-projects/data-quality-framework/testing-matrix-template gives the client statistically principled discipline over data-model outputs. But MAC alone doesn’t ensure the business compounds — it ensures the data-quality practice compounds. A client with beautiful MAC coverage on the wrong agent (bad Demand leg) or over-invested on a low-leverage workflow (bad Capital leg) just fails more precisely. The coaching curriculum should say this out loud: “here is what MAC does not do.” Candidate drip-course module: “When MAC is irrelevant.”

2. The agent-deployer role lives inside the operations leg. Per 2026-04-14-levie-agent-deployer-role-jd, agent-deployers instrument AI workflows and run MAC-style evals. That’s George Koenigsaecker at Jake Brake — a brilliant Lean operator inside a capital-allocation machine. Without the Rales and Allender choosing which factories to buy, Koenigsaecker’s Lean would have improved one division and gone nowhere. The RDCO equivalent: the enterprise needs an allocation thesis about which agents to build before the deployer’s discipline compounds. Note for 2026-04-14-levie-agent-deployer-role-jd: this is the bounding caveat on the role.

3. State-ownership needs a Demand/Capital companion. The ../04-tooling/rdco-state-ownership-architecture establishes what the client owns (vault + skills + data + the harness). This article is the reminder that owning state doesn’t tell you which state to build. Chin’s Culp-era lesson — high-aftermarket, high-gross-margin businesses — is an explicit Demand-selection criterion. RDCO’s consulting intake should include a framing conversation: where is this agent portfolio earning vs. cost of capital? If the client’s agent bets are Google-like (30%+ return on labor hours saved), MAC is polish. If they’re marginal (barely beating the hurdle of not building the agent), MAC is the difference. Candidate addition: a “limits section” in the architecture doc naming the operations-leg scope.

4. phData/MG competitive framing: most data consultancies are pure operations plays. phData sells better data pipelines; MG sells better data governance. Both are operations-leg offerings and can die the Danaher-adjacent death — excellent execution on a service whose demand curve has shifted. RDCO’s honest differentiation: we ship operational discipline plus a framing that names what this discipline does not do. The “return sweet spot” lens is a genuine wedge — a prospect earning Google-margins on their AI investment doesn’t need us; a prospect bleeding out below cost of capital can’t be saved by us; the prospect in the $1.03–$1.06 zone is exactly where MAC + harness compounds into real equity value. That’s a qualifying question RDCO’s sales conversations should ask out loud.

The meta-lesson for RDCO voice. Chin publishes the caveat immediately after the thesis. That sequencing is the move — claim the principle, then publicly patrol its limits. Sanity Check newsletter voice should do the same: every operational-excellence claim carries a “here’s where this breaks” paragraph. The Danaher arc is a perfect open-source case for a future Sanity Check issue — “Your MAC practice is Jake Brake. Who’s your Rales brothers?”