Design of Experiments (DOE) is a structured statistical method for planning tests that estimate how process inputs affect a measured response, including interactions between inputs. An interaction occurs when the effect of one input depends on another input’s setting.
In a sealing process, a team might test temperature and pressure at two levels each, producing four treatment combinations in a full factorial design. Testing every combination allows the team to assess whether higher temperature improves seal strength differently at low and high pressure. One-factor-at-a-time testing does not resolve that interaction across the proposed settings. The operating decision concerns a combination of settings, so the experiment must support that choice.
A DOE plan defines the response and its measurement procedure before allocating production runs. Replication provides an estimate of experimental error. Randomized run order reduces the risk of confusing input effects with time-related changes, while blocking accounts for known nuisance variables such as material batches. Four treatment combinations therefore do not necessarily mean four production runs. Fractional factorial designs reduce the initial run count, but aliasing can prevent separate estimation of effects and require follow-up testing.
Within DMAIC, DOE commonly supports Improve-phase decisions after root cause analysis has identified candidate inputs. Gage R&R helps assess measurement suitability when operator differences or repeated readings may affect results. Confirmation runs test the proposed settings against the model’s predictions before implementation. The control plan then documents the confirmed settings and tested range, while control charts monitor subsequent process behavior.
