Considerations for Data Sampling

Accuracy during process evaluation is critical to the success of lean Six Sigma projects. In order to generate accurate, reliable data, attention must be given to data sampling techniques. These considerations should be incorporated as part of the evaluation plan.

In drawing up a control chart for either lean manufacturing Six Sigma or lean Six Sigma for service projects, three issues in particular require consideration:

  • Sample size
  • Sampling frequency
  • Sample collection techniques

Considerations for Sample Size

  • Sample size should be realistic given financial and practicality constraints.
  • Larger samples are better indicators of out of control conditions and excessive variation.
  • Overly large samples risk including special cause variation and are not representative of true conditions.

There will be some variation to fit the environment under study, but typically a subgroup size of 5 is used for ease of computation as well as validity of results.

Considerations for Sampling Frequency

It is critical to use adequate sampling frequency generate samples that are useful in identifying sources of process waste. Samples collected too infrequently to be accurate are a common source of error in lean Six Sigma process evaluations.

For best results, the use the following guidelines:

  • Stability - Processes that exhibit erratic performance and have not been analyzed previously should be subject to frequent sampling. The increased frequency increases improvement opportunities. As a process grows more stable, however, the frequency of sampling may be reduced.
  • Process Events - Complex processes that include multiple events such as material and tool changes also require frequent sampling. In this case, however, data sampling should coincide with component event. This sampling pattern provides an accurate picture of processes' performance as well as the efficiency of the events involved in it. Unless the appropriate sampling frequency is used, the lean Six Sigma evaluation may not identify all sources of waste.
  • Cost - The process of sampling incurs costs in one of two ways, either in time used to take a sample or in material loss as a result of destructive testing. Before determining the appropriate sampling frequency for any lean Six Sigma project, the total anticipated cost of sample should be considered. However, sampling frequency that is reduced too greatly as a cost-saving measure will yield a body of inaccurate, unreliable data. The best solution is to weigh budgetary concerns against the realistic demands of accurate sampling practices.

Considerations for Sample Collection Techniques

The manner in which data is collected can affect the quality of information gathered.

A body of measurements should be collected within as close a time frame as possible to diminish the chance of variation. Consecutive sampling often proves to be the most reliable technique.

The timing of sampling sessions does not have to be rigidly set. It is unnecessary to sample using a single time interval or at the same time every day. Some variation actually generates a better quality sample. Allowing some randomness in sampling times minimizes the effects of external factors such as shift changes.