Skip to content
Six Sigma | Part 2 of 10

Choose the Problem Before You Choose the Tools

Most Six Sigma programs fail on project selection. Cost of poor quality and a simple score find the projects worth the effort.

A company trains twelve green belts and asks each one to bring a project. Six months later, four projects have stalled for lack of data, three solved problems that were already understood, and two improved processes that cost almost nothing.

The statistics were fine. The choice of projects was not.

What makes a good project

A good Six Sigma project has five properties:

  • Chronic: the problem has been there for months, not a single event.
  • Measurable: there is a characteristic that can be measured, ideally as a value, not just good or bad.
  • Valuable: the cost is known and large enough to justify months of work.
  • Bounded: one process, one team, four to six months.
  • Unexplained: the cause is not known. If the solution is obvious, just implement it.

Problems with an obvious cause belong to everyday problem solving at the line. Known mistakes that must never happen belong to error proofing. DMAIC is for the problems that resist both.

Start with the cost of poor quality

The best project list comes from the controller, not from a brainstorming session. Ask for the cost of poor quality per product family: scrap, rework, sorting, extra inspection, warranty, complaints and expedited freight.

In many manufacturing companies this adds up to 5 to 15 percent of sales. It is rarely in one report, which is why it is rarely managed.

A simple selection score

Rate each candidate from 1 to 5 on four criteria:

  • benefit, from the cost of poor quality,
  • strategic fit,
  • probability of success, including data and access to the process,
  • effort, where 5 means low effort.

Score = benefit × 2 + fit + success + effort.

Candidate Cost per year Score
Bore diameter of a pump housing EUR 84,000 23
Leak test rejects in assembly EUR 61,000 19
Porosity in die casting EUR 112,000 18

Porosity costs the most but ranks third: without the foundry supplier on the team, the chance of success is low. The score does not decide alone. It makes the trade-off visible, and the champion decides.

The problem statement

Once selected, the problem gets a statement that contains a number and no cause: “The bore diameter of housing PH-24 has been above the upper limit at a rate of 12,300 ppm since January 2026, found at the final check.”

Words like “because”, “due to” or “lack of” in a problem statement are a warning. They contain a solution that has not been proven.

One project per belt

Green belts run projects next to their daily work, typically 20 to 30 percent of their time. One project at a time is the rule. Two parallel projects usually mean two late projects.

Selection is the cheapest point to improve a program. A good project with average statistics beats a poor project with perfect statistics every time.

Newsletter

One Lean idea.
Every two weeks.