Take Time to Plan Before Collecting Data

Posted by on Jan 5, 2011 in Uncategorized | 0 comments

The first step in the data collection process is to create a plan that will guide you in collecting consistent, reliable, and repeatable data.  Such a plan has four key elements:
What are you going to collect?  As defined by a clear and easily understood definition of what you are measuring.

How are you going to collect the data?  Clear and written procedures for how to collect your data, what tools you will use, what good versus bad data looks like, how to read the gage or measurement tool, and what forms or templates you will use to capture the data.

When will it be collected?  Detailed descriptions of what times the data will be collected.  Will it be every item or will it be by some random sampling technique?

Who will collect it?  Description of who the person is who will actually do the data collection.
Focus your data collection activity on the metric(s) that your customer is most concerned with and any secondary metrics that may apply.  For instance, if the main focus is to improve accuracy, or eliminate errors, a secondary metric that may be influenced is time.  This might be an example of time being negatively affected by your efforts on improving the primary metric, accuracy.
Most often, you won’t be able to collect data on every single action, process step, product, or transaction and you will be taking random data samples.  When doing so, there are two decisions to make.  The first is how to sample or what process to follow and the second is the sample size you’ll need to be large enough to draw the conclusions you’ll need.  Here are some guidelines:
Collect the data randomly.  The data should be representative of the process.

Use a systematic approach.  Take a sample every five units, or take one sample every five minutes.

Rule of thumb for sample size.  The larger your sample, the more certain you can be that it is representative of the population from which it is drawn.
Some things you don’t want to do are:
  • Collect the data based on how easy it is to collect.  Easy data doesn’t represent the whole process.
  • Be arbitrary or inconsistent.  Taking too many samples in the morning and none in the afternoon, or taking all samples on Monday and none on Friday.

Leave a Reply

Your email address will not be published.