(544a) Experimental Design of Sample Plan to Enable Data-Based Decision Making

There are hundreds of analytical labs in The Dow Chemical Company.   A main purpose of these labs is to enable process control and to ensure product quality consistency.  Another purpose is to support special improvement projects.  Lab operation needs to follow a sample plan (i.e., a document describing what, where, how, and how often to sample from the manufacturing process).  A statistically sound sample plan is crucial across all businesses in Dow because insufficient samples can lead to wrong decision making  and too many samples could be just a waste in time and cost.    

This talk will discuss experimental design of sample plan to enable data-based decision making such as product variation investigation, accepting or rejecting incoming raw materials, and deciding analytical measurement improvement activities.   Two industrial examples are given here: 1. Optimize a current sample plan to reduce number of samples and 2. Design a new sample plan to understand the physical properties of catalyst. In both of these projects, sound statistical methods were used that take into account both risk and costs. These are critical information that helps Dow to make important data-based decisions.