Return to the Table of Contents.

10.1 Collecting data

While in graduate school, I worked part time for a medical provider. On one project, I used data showing information on all the patients admitted to a particular hospital over a six month period. I first prepared summaries of the data in order to begin to understand patterns. During that process of getting to know the data, I noticed that the admissions included quite a few infants, just hours old. After puzzling over the high numbers, I mentioned the fact to my boss. He, too, was puzzled. We called the data source and learned that, of course, every new born child was an “admission” in order to enter the child into the hospital database. We didn’t bother calling the data source when I discovered that the database included a pregnant male; that record clearly had an error.

That story has three lessons in it:

• Get to know your data.
• Know how the data were collected and coded.
• All data sets have errors.

Data are crucial to fact based decision making. In many organizations, conventional wisdom may not be true, and only data, carefully collected and analyzed, can help uncover truth. In many studies I have done, I have analyzed data already collected by other. Using existing data sets saves time, but has drawbacks. Since the data were collected for a different purpose, some important types of data may not have been collected. Also, as the above example illustrates, you have to understand how the data were collected and coded. For example, data on process times in a manufacturing plant may or may not include set up time.

Data that the IE collects for a particular study may be collected as part of the normal production process (for example, measure the weight of the first 4 bags of coffee produced each hour) or may be part of designed experiment. For example, to determine whether quality problems are due to differences in machine, worker, or supplier of input material, an IE may have each combination of machine, worker, and supplier produce a set number of products.

Since collecting data, either in the production process but especially in a designed experiment, takes time and resources, the IE must think first about how the data will be used. Data on the service time to serve each customer can be used for different purposes if data are also collected on the time of day, the employee who served the customer, and the type of service the customer needed.