On Process & Efficiency
ANNOTATED
The printed text is the standard business-school account of process and efficiency. The handwriting is what a decade of running actual processes — a Made-to-Measure line, SaaS rollouts, a business from zero — did to that account.
The Nature of Process
The textbook defines a process as the connective tissue of an organisation. Where an individual performs a task, a process ensures that the task is performed the same way regardless of who executes it. This is held to be the foundation of scale: a business grows not by finding more talented people, but by encoding what its talented people know into steps that ordinary people can follow reliably.
Standardisation reduces variation. In manufacturing, variation is the enemy of quality; two units that should be identical but are not represent a defect, a cost, or a customer complaint waiting to happen. The discipline of process design is therefore the discipline of removing unwanted variation until the output falls reliably within tolerance.
Because a process is written down, it can be examined. It can be measured, questioned, and improved. A task that lives only in one person's head cannot be improved by anyone else, and disappears entirely when that person leaves. Process, in this sense, is how an organisation remembers.
Ray Kroc didn't scale McDonald's on the strength of a secret burger. He scaled it on the "Speedee Service System" — the McDonald brothers' obsessively documented kitchen choreography that let a teenager, anywhere, produce an identical meal in under a minute. The franchise empire was possible only because the process was so completely encoded that it no longer depended on any individual's talent. That's the whole idea: they franchised a system, not a chef.
Running a hundred-person Made-to-Measure department taught me this before any textbook did: the day I wrote the measurement-flow down so any operator could run it, the department stopped depending on me being on the floor. Years later, building a flooring business from zero, the first thing I did was write the job process down on day one — before we were busy, not after. — A.P.
Growth is achieved by scaling process, not effort. To grow, encode the work.
The Meaning of Efficiency
Efficiency is commonly taught as a cost function. The efficient firm produces the same output at lower cost, or greater output at the same cost, and thereby earns a margin advantage over less efficient competitors. Under this view, the pursuit of efficiency is the pursuit of profit by another name, and the two are treated as interchangeable.
The standard treatment warns against two failure modes. The first is inefficiency by neglect: waste that accumulates simply because no one is looking. The second is false efficiency: local optimisation that improves one step while degrading the system around it. A department that hits its own targets while starving the next department of what it needs has not become efficient; it has merely exported its inefficiency.
True efficiency, therefore, is a property of the whole system, not of any single part. It is measured at the point where value reaches the customer, not at the point where any individual step reports success.
When Amazon reorganised its fulfilment centres, it didn't optimise the pickers' walking speed in isolation — it redesigned the whole system so items came to the worker, measuring success only at the point the customer received the package on time. A picker moving faster inside a badly-laid-out warehouse is the "faster step in a slower system" this rule warns against. The discipline: Amazon judges the whole stream to the doorstep, never one heroic step in the middle.
In home services this rule gets personal fast. Every redone install steals a tradesperson's day, a customer's patience, and the next job's start date. When I stopped measuring my crews on "jobs started" and started measuring the whole stream to a happy handover, the business got calmer and faster at the same time. — A.P.
Optimise the system, not the step. A faster step inside a slower system is decoration.
- Process copies judgement, not motions. If it only copies motions, you built a script, not a system.
- The process is the org's memory. Undocumented work is a company you rent back from your own staff.
- Efficiency is respect, not just cost. Waste is always paid for by a person downstream.
- Optimise the handoff, not the step. Local heroes who export their mess aren't efficient — they're loud.
The Transitive Trap
The transitive property is one of the most seductive tools in a strategist's kit. It lets you build a chain of reasoning from a starting move to a desired outcome. If better onboarding raises activation, and higher activation raises retention, then investing in onboarding must raise retention. The logic feels airtight. It has the clean inevitability of arithmetic.
b = c (measured by ruler Y)
∴ a = c ← only if X = Y
lose the proof ✎
Consider a concrete case. A team observes that customers who use a feature are more valuable, and concludes that driving more customers to the feature will raise customer value. But "customers who use the feature are valuable" was measured by observing existing behaviour, while "driving customers to the feature raises value" is a claim about causation. The first is a correlation on one population; the second is an intervention on a different one. The equals sign between them was never earned.
The failure is rarely in the individual claims. Each link, examined alone, may be perfectly true. The failure is in the join — the assumption that because two relationships were each observed to hold, the metric defining one is the same metric defining the other. Change how you compare A to B versus how you compare B to C, and the transitive conclusion collapses without a single link being false.
A startup notices power users engage with a certain feature, and that engaged users renew. Chain the logic — push everyone to the feature, engagement rises, renewals rise — and pour a marketing budget into it. Renewals don't move. Why? "Power users like the feature" was measured on people who self-selected into it; "engagement drives renewal" was measured across a different, broader base. Two true observations, two different populations, one imaginary bridge. This exact mistake has burned more well-funded growth teams than any competitor ever did.
I now audit every client growth model for exactly this trap. In consulting, the chain usually reads "more content → more traffic → more sales" — three different rulers, three different populations. The kindest thing I do for a client is make them name the ruler on each arrow before they spend a dollar on the plan. — A.P.
If A drives B and B drives C, then A drives C. …only when both links share one ruler, one population, and one direction of causation.
- Transitive logic is a loan, not a proof. a=b and b=c only give you a=c if the measurement behind both is identical.
- Chains snap silently. Each link stays "true" while the conclusion becomes false — nothing lights up red.
- Name the ruler for every link. How do you know, and on whom was it measured? The weak join confesses itself.
- Correlation observed ≠ intervention promised. Watching valuable users is not the same as making users valuable.