Improve
The Improve domain generates, selects, pilots, and implements changes that close the verified gap. ICGB tests whether you choose solutions with evidence and confirm improvement rather than assuming it.
Generate and select
Use brainstorming, benchmarking, TRIZ, and FMEA to widen the option set. Then score options with a Pugh matrix or weighted criteria. Include impact, cost, cycle time, risk, and ease of implementation. A solution that fixes the cause but cannot be sustained is not a good solution.
Design of experiments
Screening designs identify important factors; factorial designs estimate main effects and interactions. Randomize runs, replicate where possible, and block known nuisance factors.
Lean and mistake-proofing
5S, standard work, flow, pull, and poka-yoke reduce complexity and prevent recurrence. Prefer changes that make the error impossible or automatically detected.
Worked example
A team verifies that a curing temperature is a root cause. Three settings are tested with a factorial experiment. The main effect for temperature is large, and the temperature × catalyst interaction is also significant. The team selects the setting with the best mean and lowest variation, pilots it on one line, and compares before-and-after capability. Only after Cpk improves from 0.7 to 1.2 does the change go plant-wide. Reasoning: the pilot and capability comparison separate a promising result from a proven one.
Traps
- Implementing the first plausible solution.
- Piloting without a comparison baseline.
- Ignoring interactions and optimizing one factor at a time.
- Underestimating operator impact and training needs.
- Declaring victory from a single good day of data.