Designed Experiments

Designed experiments (DOE) are one of the most powerful tools in the Improve phase because they allow you to learn how multiple factors influence a response in a controlled, efficient way. Unlike observational data, which reflects whatever the process happens to produce, DOE gives you the ability to deliberately manipulate inputs and observe the effects. This provides clear, actionable insight into cause‑and‑effect relationships. 

The strength of DOE lies in its structure. Instead of changing one factor at a time—a slow, inefficient, and often misleading approach—DOE varies multiple factors simultaneously. This allows you to detect interactions, quantify effects, and build predictive models with far fewer runs than traditional methods. 

DOE begins with defining the factors, levels, and response. Factors are the inputs you want to study—such as temperature, pressure, or staffing. Levels are the settings of each factor—such as high and low. The response is the output you want to improve. 

The design determines how the experimental runs are structured. Full factorial designs test all combinations of factor levels, providing the most complete information. Fractional factorial designs reduce the number of runs by testing only a subset of combinations, making them ideal for screening many factors. 

Randomization, replication, and blocking are essential components of DOE. Randomization protects against bias by ensuring that uncontrolled factors do not systematically influence results. Replication provides an estimate of experimental error. Blocking controls for nuisance variables that could obscure the effects of interest. 

DOE provides clear, quantitative insight into which factors matter, how they interact, and how to optimize the process. In the Improve phase, it is one of the most efficient ways to move from analysis to action.