Knowledge Nuggets – Part 5 - Data Collection – Measure Phase

I have covered Project Charter - in my previous post Knowledge Nuggets - part 3.

The next step during the Measure phase is the data collection. I want to share some of the interesting and critical points during data collection.
Data filtering or screening:

• Important step before you work on the data analysis is to do the data screening to check the validity of the data. May be data entry error, intentional data tweaking, lack of understanding of data definition. Hence start with operational definition of the data.
• If the data is not accurate and you will draw wrong / erroneous conclusions about the process behavior.

Data needs to representative of the process. what is representative? Sample data set need to capture all the factors contributing to variations in the process. This is important to study the process behavior accurately.

• For example, if you collect data to study the height of Indian population you can’t select data only from south or north. Since the height of the population varies significantly from North to Southern part of India.

Data segmentation:

• Segmentation is important to study the process behavior. You can’t mix two different things in one data set. This will hide the process behavior.
  • For example,
    • Shift operations
    • Regions
    • Machines
    • Raw material
    • Vendors and so on.

Practical aspects of Measure Phase data collection to consider:

• Few challenges one may face during the Measure Phase especially in the case of Service Industry:
  • Existing measures may not be available
  • No data capturing system in place
  • People don’t want to capture as in the case of “error” between the maker and checker. They rectify the error and never track to find out why and what kind of errors are occurring.

This will delay the project progress to proceed further within the time plan if above things are not addressed.

• The end in mind for Measure phase is to establish Baseline Process Capability.
• For service industry the best method for establishing process capability is to adopt simple DPMO method according to me.
• As service processes, in many cases, not fitting in any of the known Statistical Distribution properly.
•  No need to get stuck to find what distribution your process fits in, which is unnecessary according to me. More specifically it will never fit in the principles of normal distribution.
• Aim is to improve the process not to get stuck Statistics.

• When there is no reasonable data available?
  • Worst case you will not have enough data to establish process capability. The team should not get stuck at this stage and invest time to establish base line. Since you have enough sense of the problem please move on to the next phase. Nothing wrong.
  • Simple saying, we have no data we can’t use DMAIC is not an excuse. This way you will get away with not resolving the problem.
  • Also, as part of the measure phase, it is a good chance to define, test, validate and establish the new data collection system.

Let me conclude the post with the following comments given by Mr. Mikel Harry on my blog " DMAIC" some time back. Well-articulated about the data collection.

I will cover in my next post on the following topics Normal distribution, Process Capability and 6 & Sigma explanation.