I had a coaching call with one of our Blended Online LSS Black Belt students this week about data collection.
The transactional process discussed involves a person working through multiple screens in a software application to complete actions. The goal of the project: to reduce the time for the process in order to increase capacity with the same staffing.
The current business practice, when a process is to be improved, is to institute a measurement process requiring that all the workers use timers, time their activities, and fill out a form. The data are then collected and provided to the operators in a personal dashboard report on how they are doing compared to the group. The BB student indicated that the workload was too high to implement such a strategy in his case because the known productivity loss due to data collection would create a need for overtime, which was not acceptable. What should he do?
The advice was to consider why data collection was needed in terms of the decisions to be made. In his Lean Six Sigma project, he needed to understand the baseline performance. The group supervisors would like to know how each person is doing so that they can work with them, but that was not a need of the project asyet. This discussion led to a decision to sample randomly each day to collect information instead of requiring all participants to monitor their time.
In our courses we call this infrequent sampling, because the goal is not to judge the real-time quality of the process as seen by the customer but to judge the overall health or performance of the process when you consider all of the routine variation sources as a group. There is no effort to collect the data to quantify or analyze the different sorts of variation sources. Even sampling one transaction a day, if randomly selected, will provide a good estimate of the overall process as it is managed by the business. Over a number of days, this infrequent sampling plan should have captured the differences in people, differences in days, differences in types of transactions… which are all you need for a baseline assessment of a process. This sampling plan would not impact the work capacity either.
Now if the later phases of the improvement project show that the biggest issue is the difference in workers, then a more comprehensive data collection plan might be needed, but at the beginning you do not need it.
Common issues with Lean Six Sigma Projects: Many belts believe that they have an idea of the causes that will be found in a project and attempt to collect the baseline data with all of the demographics related to their expected cause list. Adding all the causes to the data collection goal will always require a greater amount of data to be collected in order to provide a representative sample for each demographic. Many belts say that it is worth the extra effort to save time later, but my experience has shown me that the team process will usually identify causes that the BB did not consider at the beginning, which will end up requiring a second big data collection. Why risk two intrusive data collections? Keep the eye on the data needs; only collect what is needed to make the next decision.