ARTICLE · Research

Business process optimization methods: how to choose yours and not fall for fads

Nikita Nechaev · 16.07.2026 · 6 min

Entire libraries have been written about process optimization, and almost all of that literature is built the same way: one method — Lean, Six Sigma, reengineering — is taken and declared universal. In practice the opposite holds: the method that saved a manufacturing company sinks a service business; the approach that worked for a corporation breaks a company of eighty people. What follows is an honest overview of the main methods with their limits of applicability, and a framework for choosing one to fit your situation rather than whatever business Telegram channels are writing about right now.

First, an inconvenient truth: the method is not what matters most

Over the past few decades, management thinking has produced a good dozen schools of optimization, each with its own certifications, gurus, and conferences. Yet failed optimization projects fail in the same way regardless of the school chosen: the change never reaches the people who actually work in the process. My main takeaway from practice sounds boring: a mediocre method implemented together with the people who run the process always beats a brilliant method imposed by decree. So the choice of method is a second-order question. It still matters, though, because the wrong tool can discredit the very idea of change for years to come.

The five schools — and what's inside them

Lean: remove the waste. An idea from the Toyota Production System: walk the process through the customer's eyes and strip out everything the customer does not pay for — waiting, redundant approvals, rework, a piece of paper being ferried between departments. Its strength is the low barrier to entry: a team finds its first improvements within a week through simple observation. Its weakness: Lean improves the existing process and never asks whether the process should exist at all. Ideal as a first step and as a culture; insufficient when the process is conceptually obsolete.

Six Sigma: remove the variation. A statistical approach: not "faster" but "more consistent" — so that every run of the process produces the same result. Its native habitat is mass production and large volumes of data, where a deviation of a fraction of a percent costs millions. My view, which certified Black Belts will not like: in the average mid-sized Russian business, Six Sigma is almost always overkill. Where a process runs a hundred times a month rather than a hundred thousand, the statistical machinery consumes more than it delivers. Take one thing from it — the habit of deciding on data rather than opinions.

Theory of Constraints: find the bottleneck. Goldratt's approach: in any system there is one constraint that determines the throughput of the entire flow — and improving anything other than that constraint is pointless. This is the most underrated framework for mid-sized business: it outlaws management's favorite game of "improving everything a little" and forces you to concentrate resources at a single point. Its weakness is that the method says nothing about how exactly to relieve the constraint once it is found: it answers "where," not "how." It works beautifully in tandem with Lean: TOC finds the point, Lean fixes it.

Reengineering: tear it down and build anew. This is when a process is not improved but designed from a blank sheet, as if the company did not yet exist. It delivers severalfold gains — and carries severalfold risks: people, habits, and accounting systems resist a total rebuild far more fiercely than they resist gradual change. It is appropriate in two cases: when the process is broken beyond repair (patching the holes costs more than rebuilding) and when the business model has changed while the processes are left over from the old one. Launching reengineering "as a precaution" is a reliable way to lose a year and a team.

Automation: hand it to the machine. Strictly speaking, not an optimization method but an amplifier of whatever already exists. Automated chaos is chaos that runs faster and costs more. The rule I repeat more often than any other: simplify first, automate second. A separate trap of recent years is "let's implement AI" as an end in itself: the tool gets chosen before the problem is understood. It is the equivalent of buying a lathe and then looking for something to machine on it.

A framework for choosing: four questions

Choosing a method comes down to four questions about your situation, not about the methods.

How well understood is the problem? If you know where it hurts, nature has already done the TOC work for you — start with Lean at the point of pain. If you don't, begin with diagnostics and measurement rather than a method (how to run that diagnostic is covered in our step-by-step optimization algorithm).

What does a process failure cost? Where a failure is expensive — medicine, finance, hazardous production — you need stability and control of variation: borrow the discipline of Six Sigma. Where a failure is cheap, iteration speed matters more — Lean.

Is the process obsolete, or merely clogged? If it is clogged, clean it (Lean, TOC). If it is conceptually obsolete, design it anew (reengineering) — but only that one process, not the whole company at once.

Do you have data? Statistical methods without data are theology. If your records cannot even tell you the cycle time of a process, any choice of method starts with the same step: two weeks of manual measurement. It is boring, which is why the step gets skipped — followed by surprise that "the method didn't work."

What not to do

Do not implement a method wholesale, "by the book" — take individual tools to match the task; certification is for consultants, not for your business. Do not start by buying software — process platforms help mature processes and harm chaos. Do not optimize everything at once — a portfolio of ten sluggish improvements always loses to a single one carried through to the end. And do not confuse activity with results: the only honest measure of optimization is a change in money or time, recorded before and after. Which metrics to track for that is covered in our article on measuring operational efficiency.