In my last post I invited comments on to help a reader launch ShortRun SPC in his company. I got some great responses, all of them unique and offering a slightly different perspective. They were scattered across a couple of sites and pages, or were buried in the comments section. So I thought it would be useful to consolidate all of them in one place.
Here is the original question:
Hi, I’m working in a short run / high mix products metal mechanic process.
Your articles are very interesting. Part of my job is launch SPC. So I would like to have a good start. In the past, only a few trials with XR charts, that were not good enough.
Any help and suggestions with Short Run SPC is welcome.
My response:
My advice is to start small and keep it simple. This is probably not what youre expecting from a data head who brings a comprehensive enterprise SPC Software solution to the table. But you need to build trust and confidence that SPC, especially ShortRun SPC, bring value to your organization. Pick one process and one key metric to track. It should be one where youre pretty confident that you dont need to code out the Range chart. That will help you keep the math simple, and make the overall effort more explainable to everyone.
I suggest you commit to charting this process manually for several days or a week. Start with a simple ShortRun control chart and subtract the target from each value you record so that your target line is zero. When I say chart this manually, I mean (gasp) use a pencil and a paper control chart. Write it down. It will take longer, but in the end youll understand the process better and youll be able to defend whatever conclusions you come to. Once you have the chart, treat it like you would any other control chart and see what it has to teach you about the process.
To this, Larry C added:
Pick a champion that wants to do this type of SPC work. Make sure they have enough time to make it a top priority. Have weekly update meetings to ensure it feels like an important project. Give out kudos for everyone taking the initiative to get involved and make it happen. Meetings on the subject must happen the same time every week rain or shine regardless of those in attendance. (translation this is important) Widely publish the benefits of the short run SPC work.
And Mark L added:
Paramount to making short-run SPC work beneficially is thinking PROCESS rather than product. Known differences [typically, product differences] are coded out of the data via short-run techniques. Most situations where short runs are conducted occur at stations or machines that are in operation long-term. Improving that long-term process is the real goal; short-run processes will improve as that happens.
I agree with Larrys comments above regarding the need of a champion wanting to do this type of SPC work. An important part of this type of SPC work is thinking process and helping others to think process allowing the data to guide.
In support of what Evan wrote regarding starting with one characteristic and learning how it can best benefit, I recall Dr. Deming answering a plant manager who had stated they had 130 active SPC charts and was wondering what the appropriate number might be for a plant that size. The answer was, ONE! the one you are using. Always rewarded is preliminary planning and honestly answering, What should we measure and track? [In the early days, too frequently the question was asked, “What can we measure and track?”] (Empahsis added.)
I would start by listing and prioritizing the processes (not products) that are common to the plant. For the highest priority process, answer the question, What should we measure and track? This question may also be asked as, What metric would we most like to improve?
And then Dale W added the following comment over at LinkedIn:
I also was faced with this scenario. After SPC implementation and much data collection we had two problems. 1. We had too many run charts to review and act on due to high mix. These were parts that were ran frequently, thus all the charts. Not the same problem posed here but follow me please. 2. We now have charts that did not have enough data points to even set gate limits. Worst of both worlds! Knowing we had to do something, we moved to Short Run SPC. So to try to give helpful info, this is what we did. Determine families of these short run parts. Are they ran on the same equipment? Are they of the same material? Try to group the parts based on this type of criteria with target offset.
The naming convention for these “families” is important. Now we don’t have a data collections session for pn x, we have sessions for example called “Molding Polypropylene family” As long as its is produced from the same material, we can combine data from other Injection Molding machines, get gates set and monitor the process! Remember problem number one was so many charts on the floor we couldn’t manage them all. This problem is now gone also due to short run SPC.
And the last word goes to Bruce S, on an earlier post:
My only comment on the standardized piece is that you consider using the control chart that adjusts the control limits based on each subgroup size, so that the chart looks like a skyscraper cityscape. But youve probably already considered this, and have gone the standardized route for some reason(s) which we can discuss if youd like.
I really appreciate the diversity and breadth of these answers. But what do you think? What is the most important aspect of this advice? What do you appreciate most? How could we make it even better? Use the ShareThis button below to mark this page, or leave a comment, schedule a conversation, or call 800-958-2709.
