nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Using SPC but with variable target/USL

jpollock

Hi, I am looking for some advice using SPC to monitor Actual Yield vs Planned Yield.

Plotting the SPC to measure Actual Yield (%) alone is straightforward, and normally when there's a single fixed value for an USL or a Target, this is also easy to add. However, for each month there's a Planned Yield value and this varies from month to month. The Target or USL seem to require a fixed value - so what's the best way to represent Actual Yield vs the Planned Yield in an SPC chart?

If there's a better way than SPC, such as standard graphing, then also keen to hear. Appreciate this would be fairly easy to do in Excel if that's the recommended way.

Accepted answers

All comments

P_Bartell

Here's my thought: Create a variable called "Yield Residual" or something that is descriptive to your audience where the plotted variable is just the Actual - Planned value. You can use an individuals and moving range chart to monitor the stability of the residual value.

statman

Sorry for my possible misinterpretation of your question...Here are my thoughts:

1. SPC has nothing to do with target values or spec. limits. It was invented by Shewhart to understand consistency/stability of the within subgroup sources of variation and then, if appropriate, compare those sources to the between subgroup sources of variation to determine where to investigate.

“The engineer who is successful in dividing his data initially into rational subgroups based on rational theories is therefore inherently better off in the long run. . .” Shewhart, Economic Control of Quality of Manufactured Product

2. SPC is NOT a monitoring tool! It is used to answer specific questions about the process. This does not mean you can't plot the data in a time series, but that is not SPC. I believe you should understand the consistency and stability of actual yields (and if it were me I would want to understand what affects yield and how to reduce the variation in yields). To do this requires first a set of hypotheses regarding what factors affect the yield and then directed sampling to separate and assign the causal factors (or perhaps DOE to speed the investigation up). You should also seek to understand the variation in planned yields (is this just a negotiation or are there factors driving these changes?)

I recommend reading Don Wheeler's book: Understanding Statistical Process Control

jpollock

Hi there,

Yeah, this method had crossed my mind as a potential way to go forward. However it means that if people are still interested in the raw value then there's another graph needed. Thanks though!

P_Bartell

Then any overlay plot run chart is just fine. You can overlay a spec limit as you desire. Should be easy enough to create in Graph Builder...but as @statman points out, you'll not be practicing SPC but just what I'll call 'watchful waiting' of the processes performance over time.

jpollock

Hey,

I'm not sure I understand you fully. In JMP, there are options to add a Target and/or Upper and Lower Spec limits to the SPC chart. That doesn't sound like something that would be possible in JMP if people thought they had nothing to do with SPC charts.

SPC, to me, is a way of tracking process stability and performance, giving you closer to real-time indications of when any given process goes out of control or there's been a step change.

statman

I'm not sure how to respond to your comment...No offense intended, but I would again recommend you get a better understanding of SPC before haphazardly applying it (read Shewhart, Wheeler, Neave or Deming).

Control limits have no relationship with specification limits. Usually (if not always), they are derived independently from different sources of "information". Control limits are a function of the data, how it was collected and how it was summarized. They do not answer the question about whether the product will perform as intended. Spec. limits are intended to indicate when product or process performance is compromised (Ignoring Taguchi's Loss Function ideas regarding variation around a target value). It can be confusing (and in some cases completely wrong, like in the case of the X-bar chart) to add spec. limits to control charts as the interpretation regarding control limits is completely different than spec. limits. Just because you can doesn't mean you should.

A point outside of the control limits on a range chart indicates special (Deming) or assignable (Shewhart) cause variation assignable to the sources of variation acting within the subgroup that was used to calculate the range. A point outside of the control limits on an average chart indicates the sources of variation between subgroup have more influence on the charted metric than the within sources. Neither of these uses has anything to do with specifications

jpollock

Hi,

It's fine. I understand the meaning of SPC and control limits and how they're derived from standard deviations, and I know CLs have no relationship to upper and lower spec limits or targets. Fundamentally we're using them to be able react to process going out of control (special cause) or displaying large degrees of variation. The intention is, as you've implied, to then explore the reasons why and drill down based on contributing factors.