What is DMAIC? The 5-phase Six Sigma framework explained
DMAIC stands for Define, Measure, Analyze, Improve, Control. It's the structured problem-solving framework Six Sigma uses for any process that's underperforming. If you've been handed a chronic quality problem, a yield miss, or a process variation issue, DMAIC is usually the right tool.
Define — scope the problem
You write a problem statement in dollars, defects, or hours — not adjectives. You scope what's in, out, and deferred. You agree on a primary metric and a target. The tollgate out of Define is a charter your sponsor signs.
- Problem statement with numbers (baseline, target, timeframe)
- Scope (what's in, what's out)
- Primary metric + data source
- Team and sponsor committed
- SIPOC complete
Measure — prove the baseline
You validate the measurement system (Gage R&R for continuous, Kappa for attribute). You pull 30-90 days of baseline data. You segment by shift, line, SKU, operator, material lot. If your measurement system can't measure, nothing else matters.
Analyze — find the root cause
Fishbone (Ishikawa) across the 6Ms: Manpower, Machine, Method, Material, Measurement, Environment. Then hypothesize and test. A root cause isn't confirmed until data from the process proves it. 5-Why chains are cheap to build and expensive if you stop too early.
Improve — pilot the countermeasure
Pick the countermeasure that attacks the confirmed cause with the highest impact-to-effort ratio. Pilot at limited scale first. Measure the effect against the baseline. Only scale what the pilot proves.
Control — make the gain stick
Every project that didn't hold died here. Control plan per CTQ: spec, method, sample size, frequency, responsible role, reaction plan. Standard Work updated. Training completed. 30-60-90 day sustainment checks scheduled.
When DMAIC is the wrong tool
DMAIC is rigorous and takes weeks. If the problem is new (no history), use design (DMADV / DFSS). If the fix is obvious and cheap, use Process Improvement (PDCA). If the goal is flow, use Lean. DMAIC wins on chronic, data-available, process-variation problems.