The Importance of Sampling Strategies for Effective SPC Analysis

Steve Wise
By Steve Wise | December 4, 2020
Vice President of Statistical Methods

Fact checked by Stephen O'Reilly

Manufacturers—is statistical process control (SPC) viewed as a necessary evil at your organization, to be used only to remain in compliance with customer contracts? Do you collect data only to file it away? Or (and here’s my favorite) does SPC mean “Start Producing Charts” at your company? You’re not alone! Let’s talk SPC…and, more specifically, some ways that will change your view of SPC from begrudging compliance to your partner in productivity. We’ll begin with sampling strategies, in this blog.
In the previous blogs in this series, we’ve discussed ways to reduce waste, scrap, and rework, we’ve taken a look at the top advantages to using SPC in your manufacturing operations, and we’ve zeroed in on “zero defects” (and making that phrase more than just a hollow slogan, but something to strive for, at your company). One way to make the most of your SPC efforts at your organization is with sampling strategies. Let’s dive in.
Sampling Strategy 

Aiming for Higher Quality with SPC Analysis

How many of you have ever embarked on collecting SPC data only to realize that the information provided by the control charts and histograms was not very helpful in reducing costs, eliminating waste, identifying defects, or reducing variation?

As mentioned, we’ve discussed identifying and reducing defects at length. Eliminating defects at their source is the key to higher quality, increased efficiencies, more throughput, happier employees, happier customers, lower warranty costs, higher profit margins, and more…you get the point.
And we concluded that asking the right questions will lead you to where the defects are first introduced into your manufacturing processes. How you structure your data collection strategies will give you the answers you need.

Data and Questions

Data is only as good as the questions it can answer. You can put all the data you want into your database, but if it doesn’t answer the questions that are important to you, then it’s useless. The more you know about the questions that need answers, the better you’ll be able to construct the right sampling strategies.
SPC and Sampling Strategies 
How am I going to collect this data? How often I collect it? What sort of measurement device or instrument do I need to use to collect this data? And then what questions do I need to ask or answer from the data? And then based on those sorts of questions, and having all that knowledge, will dictate how we're going to put the sampling strategies together. So that's what a sampling strategy, essentially, it's just the thought process of collecting the data, and what goes into collecting data.

Getting Started

Data collection and SPC analysis are really the most beneficial when the efforts lead to solving or preventing problems. But sometimes the toughest question is “Where do I start?” Let’s do what Dr. Deming says and just start “somewhere.” (If you need a Deming fix, please read COO Doug Fair’s blog series about Dr. Deming’s 14 Points for Management…there are plenty of gold nuggets in there.)
One of my favorite places to start is looking at warranty claims or customer complaints. They are the end of the line. Where the “rubber meets the road.” In our example, using a simple Pareto chart, we can see that most complaints in our fictitious company are due to a faulty hinge on widget A. 
Analysis Pareto
And then the questions start to roll in. Do we have any functional data on that hinge? And we discover that there is, indeed, hinge data, but widget A has two hinges. The complaints are on the left hinge, but the inspection plan just points to the testing of one of the hinges. You already see the problem(s) beginning to form.
The original inspection plan was designed to answer different questions than what the warranty investigator needs.

Doing a deeper dive, the investigator discovers that the hinge is made from two sub-assemblies. There are three machines that make the components and two suppliers that do the plating. There are also seven assembly stations (employees) where the hinges are attached to the widget. And there are other sources of variation, like batch numbers and raw material suppliers.
Analysis Pareto 
Given all these inputs, a sampling strategy would need to create a test that simulates the failure mode on the left hinge. Each test result value would then need to be associated with the appropriate assembly station, the machines that made the sub-assemblies, the plating supplier, component batch numbers, and raw material suppliers.
So, we collect some data and then use InfinityQS’ SPC tools to analyze the data for any patterns. Given the right information, something in that data will expose the path to where we want to go to next.
Data Collection and Analysis

Being Proactive with SPC Analysis

To really understand what we’re talking about here, let’s look at some other examples.
  • If fill volume is critical, then you need to make sure the fill data that you collect is associated with the specific filling machine in question AND the appropriate fill nozzle.
  • If weight of an injected molded part is important, make sure to capture the press number, mold number, and cavity number, plus any inserts. As well as the process input parameter set points—like flow rate, temperature, and pressure.
Many times, the important factor that will crack the case is confounded in the data. That’s why it’s so important to do your homework ahead of time, really strategize on what data you collect and how you want to collect it.
  • If there are different measurement gauges used to verify a lot of material, the problems might be with the gauges, but if the gauge ID is not associated with the value, then the data are confounded. You’ll never know which gauge to look at if the gauge ID is not included in your sampling strategy.
  • If surface flatness uniformity is important, make sure multiple flatness measurements across the surface are collected. If flatness problems can manifest themselves on certain zones on the surface, make sure the precise zone location is associated with the flatness measurement.
Overall, having the numbers is great, but capturing the conditions surrounding those numbers is what can turn just another control chart into the “best thing ever.”
To read other blogs in this series:
Take advantage of the technology at your fingertips today: contact one of our account managers (1.800.772.7978 or via our website) for more information.


InfinityQS Fact Checking Standards

InfinityQS is committed to delivering content that adheres to the highest editorial standards for objective analysis, accuracy, and sourcing.

  • We have a zero-tolerance policy regarding any level of plagiarism or malicious intent from our writers and contributors.
  • All referenced articles, research, and studies must be from reputable publications, relevant organizations, or government agencies.
  • Where possible, studies, quotes, and statistics used in a blog article contain a reference to the original source. The article must also clearly indicate why any statistics presented are relevant.
  • We confirm the accuracy of all original insights, whether our opinion, a source’s comment, or a third-party source so as not to perpetuate myth or false statements.



Never miss a post. Sign up to receive a weekly roundup of the latest Quality Check blogs.