Rational Subgrouping in Sampling for a Better Understanding of Process Behavior
When it comes to sampling subgroups, more is not always better.
In our recent article Sampling Frequency in Manufacturing: Too Much of a Good Thing? we discussed why sampling frequency is critical to understanding process behavior and how sampling too frequently can make a control chart useless. In the manufacturing world, too much of a good thing (data) tends to bring confusion, lack of clarity, and misunderstanding.
A good sampling plan allows for a better understanding of process behavior, which enables you to effectively monitor and improve the process. If the sampling plan isn’t adequate, then there is no true understanding of the process behavior, and we’re just wasting our time.
Sampling at the Proper Frequency
Sampling frequency depends on how quickly a process changes, and in the previous article, we discussed our tendency to sample too often. We also tend to collect too many samples within a subgroup, which similarly leads to issues with our control charts.
Sampling subgroup size depends on how many products can be produced without process adjustments or raw material changes. However, we sometimes ignore the process and follow our natural tendency, assuming more subgroups or a larger subgroup size is better.
One Works Just Fine, Thanks
Sometimes ONE subgroup is the answer! When a process requires adjustments, changes to raw materials, or a unique setting for each product, a subgroup size of one should be used. Attempting to group samples produced under different conditions would include signals of process change that could make the control chart useless. The goal is to understand process signals between subgroups while the subgroup itself includes the noise or normal variation of the process.
Likewise, we should use a subgroup of one when sampling from a homogeneous batch. For example, a chemical batch could return the same sample value when measured repeatedly, or a product could return the same value. Using more than one sample per subgroup would show very little or no noise or variation within the subgroup, resulting in an overly sensitive control chart.
One sample per subgroup is also correct when only one value represents the condition monitored, such as a daily yield or weekly defect total.
But Sometimes One is the Loneliest Number
Subgroup size should be more than one when several parts can be produced without any process adjustments. Odd sample sizes are recommended as they have a natural center point or median.
We must be aware that even when subgroups of greater than one sample are possible, a higher number of subgroup samples will increase the sensitivity of the control chart and increase the risk of a false signal. The time required for sampling and measurement should also be considered to avoid the high cost of oversampling.
Taking a Rational Subgroup When Sampling
After determining when to sample and how many products to sample, we must ensure that our sampling itself is rational. Each sample should be random. A random sample is a sample in which every product within the population has an equal chance of being selected. That might sound simple and obvious, but we must be careful not to introduce bias into our sampling and risk giving some products a higher probability of being selected than others.
Let’s say we are baking three pizza pies in a deep brick oven. Checking only the first pie, closest to the door, doesn’t give us any information on how the other two pies are cooking. If the oven’s temperature isn’t consistent throughout, then this type of sampling bias could result in a couple of burned—or undercooked—pizzas.
Considerations for Designing a Sampling Plan
The following are some reasonable considerations for designing a sampling plan:
Who will be collecting the data?
Evaluate the abilities of the operator collecting the data. How much time does the operator have? Does the operator have adequate resources to collect the data?
What is to be measured?
Focus on key characteristics. Always remember that it costs money to sample. We should focus on the characteristics that are critical to controlling the process.
At what point in the process will the sample be taken?
The sample should be taken at a point in the process that allows the data to be used for process control.
When will the process be sampled?
Samples must be taken often enough to reflect shifts in the process. A good rule of thumb is that two subgroups are taken between process shifts.
Why is this sample being taken?
Will the data be used for product control or process control?
How will the data be collected?
Will samples be measured or evaluated by hand, or will the data be retrieved from an automated measurement source?
How many samples will be taken?
The sample quantity should be adequate for control without being too large and creating control limits that are too sensitive.
A Better Sampling Plan Starts Today
If you realize you’ve been oversampling for a while now, take heart: it doesn’t have to continue. When it comes to creating a better sampling plan, there’s no time like the present.
We don’t want to oversample or sample too frequently, but we need to see the data to start the improvement process. Don’t spend too much time or worry about creating the “perfect” sampling plan. You’ll benefit from collecting data and understanding your process with adequate, rational sampling—so please don’t postpone the gains!
Take advantage of the technology at your fingertips today. Contact an InfinityQS account manager by calling 1.800.772.7978 or through our website.