Determining Primary Metric
- October 1, 2018 at 8:46 pm #56103
I am in the Define phase of my DMAIC project and I am having a difficult time determining a primary metric to use. Management would like to see variation decreased in a process and they aren’t too specific other than that. I thought it would be as simple as looking at the data and finding the amount that fall outside of the specification limit and calculating % yield or % defective except very little actually fall outside of the specification limits yet there seems to be quite a bit of variation. Is this project worth pursuing since there is very very little outside of the specification limits? Could I somehow include control limits as my primary metric instead? Use Ppk or STDEV? I know I am asking a lot of questions here so any help would be greatly appreciated.October 2, 2018 at 12:55 am #203075
Consider what your primary metric is for. It’s how you’re going to track and confirm changes to the process. What is the problem? Parts out of spec? Excessive variation? You say very little is out of spec, but that is still parts being made incorrectly. So we can extract from that parts defective, or yield%. We can also as you say monitor the STDEV, or if it’s quite tight just Variance, which is STDEV^2, handy if you’re monitoring small changes. You have also identified Ppk, which could be handy but I personally wouldn’t use it as the outward facing KPI/Metric just for understanding, I like it to be as basic as possible to everyone involved. Of couse monitor it yourself though, it’s a useful metric.
You have to define the problem in some more detail. Your KPIs should fall out of that to be honest, if you’ve drilled down enough. Process has high variation > Some parts out of spec > Costs x every month/year > Parts out of spec because y reasons > Of these reasons y is the main defect/issue > It specifically costs xyz every quarter
From that kind of basic analysis you can then lay out your project to say “In the period Q1-Q3, plastic cracking on the acme parts has caused $1000 worth of scrap, including rework time and missed deliveries in the shipping dept.” along those lines. We can then see that by following that process what I would need to do is monitor the quanitity and probably severity of plastic cracking (or whatever the defect is) on those parts and look to remove it.
It’s difficult to be any more specific because I don’t know anything about your process but I hope this has helped at least a bit. I’m no expert by any stretch, just trying to help!October 2, 2018 at 6:13 am #203076
Define is all about just getting focused. It seems simple to get started, usually it is not actually that simple with very broad statements of the problems you are having, just as you described.
Start with answering the question…’What problem am I trying to solve?’
Then answer the question…’What does success look like?’
If you can put together a well thought out problem statement, then you can get into defining how to measure the current state and then future state of your problem/issue as you do your DMAIC magic.
In most cases, as you have described above, your current measurements systems are not the most accurate or best gauge of your problem and solution. You may have to gather new data, or create a more robust tracking system based on what your goal looks like.
In the way you have portrayed where you are, you actually have started with Measure before you fully completed your Define work so are shotgunning the measurement vs. a sniper like focused approach.October 2, 2018 at 12:53 pm #203085
Let your graphical and statistical analysis help you create a metric with your project champion but they ought to also have a secondary metric of what else would improve with reduced variation.
You must be logged in to reply to this topic.