Six Sigma

Running Sigma Level as a Quality Management Scale

Sigma level turns a 100x defect difference into one digit. How to compute it honestly, benchmark it, and use it to aim projects and capital.

Sigma level matters because it converts quality into a scale that spans a 100x defect difference in one digit. Three sigma is roughly 66,800 defects per million opportunities; four sigma is 6,210; five sigma is 233. Each step cuts defects 10 to 27 fold, and cost follows: three-sigma operations typically carry quality costs of 15 to 25 percent of sales, four-sigma around 10 to 15 percent, five-sigma under 5 percent. That makes sigma a money scale, not a statistics trophy. A plant that knows it runs 3.6 sigma on a key line knows, within a few hundred thousand dollars, what improvement is worth before chartering a single project.

The mechanics run defects to DPMO to sigma. Count defects, units, and opportunities per unit, where opportunities are the places a defect could occur, counted consistently. If 850 units with 12 opportunities each produced 43 defects, DPMO is 43 divided by 10,200, times one million, which is 4,216, roughly 4.1 sigma with the conventional 1.5 shift. The Sigma Level calculator handles the DPMO conversion and the sigma lookup. The trap is opportunity inflation: doubling counted opportunities from 12 to 24 halves DPMO and adds about 0.2 sigma while nothing improved. Freeze the opportunity count per product in writing and change it only with a documented recount.

Benchmarks: average manufacturing processes run 3 to 4 sigma, 6,200 to 66,800 DPMO. Strong plants operate at 4 to 5 sigma. Airline baggage handling runs near 4 sigma; commercial flight safety runs beyond 6. Within a plant, machining on capable equipment often measures 4.5 to 5.5 sigma, manual assembly 3 to 4, and administrative processes like order entry frequently sit at an embarrassing 2.5 to 3.5, which is 22,700 to 158,000 DPMO, and is why the biggest sigma projects are often in the office. Baseline each process separately; a blended plant-level sigma hides exactly the variation you need to see.

Match the levers to the altitude. Below 3.5 sigma, the work is basic stability: standard work, error-proofing, and killing the top Pareto items; teams commonly gain a full sigma in 6 to 12 months here. From 3.5 to 4.5, the work is variation reduction: SPC on critical characteristics, gage R and R under 10 percent, and driving Cpk from 1.0 toward 1.33, worth roughly 2,600 PPM per characteristic. Above 4.5 sigma, defects are too rare for chart-watching; progress comes from designed experiments, tolerance redesign, and supplier capability, and each 0.5 sigma costs more than a full sigma did at the bottom.

Failure modes: sigma theater is the big one, belts certified and banners hung while DPMO sits unchanged; measure the program in closed projects with audited dollars, and $50,000 to $250,000 per black belt project is the working range. Second, the 1.5 sigma shift confusion: short-term capability data and long-term defect counts differ by roughly 1.5 sigma, so label which one you report or two departments will fight over a phantom gap. Third, sample starvation: at 4.5 sigma, 1,350 DPMO, you need hundreds of thousands of opportunities to see change, so monthly sigma on 5,000 units is a random number generator. Aggregate quarterly at high sigma.

Cadence: monthly, compute DPMO and sigma by process family, trended over 13 months, reviewed alongside cost of poor quality so the statistical and financial views reconcile. Quarterly, leadership stack-ranks processes against a target ladder, for example every process above 4.0 within 24 months, and charters projects on the bottom three; each gets a baseline DPMO, a target, and a dollar value signed by finance. Annually, re-audit opportunity counts and defect definitions. Keep sigma off the daily boards; days belong to defect counts and first pass yield. Sigma is the strategic layer that decides where the belts and the capital go.

World-class programs hold critical processes at 5 sigma or better, 233 DPMO, keep every process above 4, and retire a measurable 0.2 to 0.4 sigma of gap per year on focus areas. Financially they bank 2 to 4 percent of sales annually from the project pipeline, audited, not estimated. Culturally, the tell is vocabulary: supervisors quote DPMO on their own lines and know whether this month's move is signal or noise. Getting from 3.5 to 4.5 sigma plant-wide is typically a 3 to 4 year program; the plants that make it treat sigma as the scoreboard and the daily FPY grind as the game.

Published 2026-07-02.