The Origin of Six Sigma is Not a Statistics Textbook, but a Repair Ledger — Motorola 1986: How Bill Smith Turned 3.4 ppm into a Corporate Strategy

In 1985, a senior engineer from Motorola's Communications Division walked into the CEO's office with a stack of data. His conclusion was straightforward: the proportion of products sold to customers that failed at the customer site was highly correlated with the number of times they were repaired during the manufacturing process—boards that were repaired were more likely to fail on-site. The labor and materials consumed by repairs were not accounted for in any specific category in the company's books.

The atmosphere in the meeting room was not good. The most widely circulated version of the story is that this presentation was seen as "pessimistic," and the presenter almost lost his job. Bob Galvin kept the materials and spent the weekend reviewing them. On Monday, he made a different decision: not to reprimand the engineer, but to set "Six Sigma" as the company's goal. In 1986, Motorola officially launched Six Sigma; in 1987, it announced its goal—to achieve Six Sigma level, or 3.4 ppm, within five years; in 1988, Motorola won the first Malcolm Baldrige National Quality Award.

This history is worth a close look by today's quality engineers, as it illustrates one thing: Six Sigma was never about statistical methods first and management movements second; it was about an unaccounted-for ledger that forced the creation of a statistical language.

1. Conclusion First: The Driving Force is Cost, Not Statistics

Understanding Six Sigma as "a set of statistical tools" is one of the main reasons for its high failure rate in Chinese manufacturing. The toolset can be purchased, trained, and outsourced, but the core issues Six Sigma addresses at its source are three business problems:

1. How much are we paying for "doing it wrong and then fixing it"? This cost is scattered across repair, scrap, expedited orders, overtime, and customer claims, with no single report.
2. Can our quality level be measured on a scale that is comparable across departments? The injection molding workshop says "defect rate 2%," the assembly line says "repair rate 5%," and the customer says "field failure rate 800 ppm." These three statements cannot be compared, and thus cannot be prioritized.
3. Where should we invest money for improvement? Should we invest in inspection (to detect issues earlier) or in the process (to prevent issues from occurring)? Without a common scale, this decision can only be made by departmental volume.

Therefore, the backbone of Six Sigma is actually "ledger + scale + project-based" three elements, and DMAIC is just a container that organizes these three elements into an executable sequence. To understand what should be delivered at each stage of DMAIC, refer to DMAIC Project Charter and Stage Gates; to see what a complete project looks like, see Six Sigma DMAIC in Practice: From a Defect Rate of 8.5% to 0.3%.

2. 1974～1986: A Company Backed into a Corner

Motorola's Six Sigma was not a strategy conceived in a meeting room; it was a response to market competition. Key milestones:

Two points that are often overlooked:

First, exiting consumer electronics did not make the problem disappear. After retreating to communications and semiconductors, the company still faced the same customer demands and cost structures. The business could change, but the defect rates in the process would not automatically change.

Second, the "10-fold" goal set in 1981 could not be implemented at the time. A goal without a measurement system can only become a slogan. The four years of inaction highlighted the value of Bill Smith's analysis—he provided a measurement method, not a new goal.

3. Bill Smith's Three Discoveries, Still Relevant Today

Bill Smith's contribution was not in statistics but in translating engineering data into business language. His findings can be broken down into three points, each directly applicable to today's production floor.

First: Defects are not detected by inspection, but are absorbed by the "hidden factory."

When a part is judged nonconforming, repaired, and then flows into the next process, the ledger cost remains almost unchanged—the materials are the same, the people are the same, and the production volume is still met. What actually happens is: part of the capacity is occupied by rework, lead times are extended, and hidden defects are carried into the final product. This work is invisible in the reports, so management does not see its scale.

The quickest way to determine if your factory has a hidden factory is to check three numbers: the proportion of repair work orders in total work hours, the number of batches accepted with concessions (special acceptance), and the number of parts that pass inspection but are complained about by the assembly line. If all three numbers are significant, it indicates a gap in the ledger.

Second: Products repaired during the manufacturing process are more likely to fail at the customer site.

This was the most impactful finding at the time, as it turned "internal quality" and "customer quality" into a single issue. The logic is straightforward: repairs often mean localized heating, secondary assembly, altered stress states, or simply "barely conforming after repeated disassembly and reassembly." The traces left by the process can be magnified in field temperature cycling, vibration, and long-term operation.

For today's quality engineers (QEs), the practical implication of this finding is: the reliability of repaired parts cannot be treated the same as conforming parts. Repaired parts must be re-verified for the specific process (not just re-inspected for appearance and dimensions), and a limit must be set for the number of repairs (common practice is no more than 2 repairs in the same area, and no more than 1 repair for key characteristics). Many companies' FMEAs do not include "repair" as a failure cause, which is a systemic flaw.

Third: Quality levels must be converted to a uniform scale to be comparable across departments and to set priorities.

This is the origin of the σ level and DPMO. Note that these are not "more advanced defect rates," but rather a way to convert different inspection methods (attribute, variable, different sample sizes) to a common denominator—defects per million opportunities (DPMO). With a uniform scale, the question of "which process to improve first, injection molding or assembly" can be calculated, not just argued.

4. The Core Conversion of 3.4 ppm and Why Not to Use It as a KPI

The core conversion relationship for Six Sigma level is as follows (the table shows the commonly used long-term drift of 1.5σ):

Two points that must be clarified:

First, 6σ in a purely short-term context is 0.002 ppm, not 3.4 ppm. The 1.5σ difference comes from observations by Mikel Harry and others on long-term process data: the process center drifts over time, and the drift is roughly equivalent to 1.5 standard deviations. Therefore, 6σ's 3.4 ppm is a number that includes the assumption of long-term drift.

Second, because it is an assumption, using σ level for departmental rankings can lead to problems. Common misuses include: treating σ level as a monthly performance indicator (large monthly fluctuations can induce data selection); comparing products with different opportunity counts (inconsistent opportunity count definitions make the denominators incomparable); and calculating capability without considering measurement system error (Cpk calculated when GR&R is out of spec is meaningless, see What is the Difference Between Cpk and Ppk and Comprehensive Guide to SPC Statistical Process Control).

The correct usage is: σ level is used for project prioritization and goal setting, not for monthly rankings. For example, prioritize 3σ workshops first, and maintain 5σ workshops; set 6σ as a three-year goal, not a quarterly performance metric.

5. Three Transformations of the Method into a Corporate Initiative

Six Sigma evolved from an analysis within a division to a global management movement through three transformations, each solving a specific problem:

First, Motorola turned it into a training and certification system. Bill Smith and Mikel Harry established courses and "Black Belt" levels, allowing the method to be replicated across different divisions. This step solved the problem of "only the inventor knows how to use it," but the side effect was that Six Sigma could become a certification movement.

Second, AlliedSignal turned it into a business metric. In 1994, Larry Bossidy directly linked Six Sigma to financial benefits, requiring each project to report a specific savings amount, with savings reaching the scale of hundreds of millions of dollars within two years. This step solved the problem of "not knowing the value of completed projects," but the side effect was that it could induce false reporting of benefits.

Third, GE turned it into a condition for management promotion. In 1995, Jack Welch launched Six Sigma, and starting in 1996, included its progress in management evaluations and required promotions to be tied to corresponding training and project completion. This step solved the problem of "lack of middle management cooperation," but the side effect was that it could become a table-driven initiative.

These three transformations correspond to three common issues in Chinese companies: only learning certification, not finance; only learning finance, not management evaluation; only learning evaluation, not the method. Without any one of these, Six Sigma can degenerate into "submitting a project PPT every quarter." This is further detailed in Practical Financial Benefit Calculation for Six Sigma Projects and Why the Black Belt Mechanism Has "Certified but No People".

6. Four Self-Inspection Criteria from the Origin

If you want to determine whether your company's Six Sigma is genuine or just a formality, use the following checklist. All four criteria must be met to be considered effective:

Additional practical reminders:

1. Fix the measurement system before discussing capability levels. Cpk without GR&R support is just a number game; when the measurement system is not up to standard, start with MSA before initiating projects.
2. Repairs must be included in FMEA and control plans. Treat "repair" as an independent failure cause in FMEA, and specify the verification items and maximum number of repairs in the control plan.
3. The financial benefit calculation must be locked in at the project initiation stage. Use financial-recognized categories (materials, labor, freight, claims, screening costs) and avoid "expected capacity savings" and other non-verifiable metrics.

7. One Sentence for Implementation

What Bill Smith did, in essence, was to link "products that fail at the customer site" with "products that were repaired in the workshop" using a data chain, allowing the boss to see how much money was being lost along this chain. Today, implementing Six Sigma in a factory starts with the same thing: first, calculate the hidden factory's costs, then discuss the tools.

The starting point of Six Sigma is an unaccounted-for repair ledger.

Knowledge code: 6.1.1

Version: v20261027

Author: QTank QTank is dedicated to providing systematic knowledge, methodologies, and practical tools for quality management professionals, helping companies continuously improve their quality capabilities.