Your name rings a bell … do I know you?
In relation to trends, it’s worth reading Wheeler “Advanced Topics in SPC” page 136 to 137. Despite the usually taught stuff on trends, Wheeler suggests they add to false alarms and that it’s better to rely on the limits. Forget about constant corrections … fiddling with the controls is what…[Read more]
You asked: “What does minimum variation really mean? ” in relation to my statement that:
World Class Quality = On Target with Minimum Variation.
In relation to variable data, good quality means much more than zero defects or fewer defects. A defect occurs when a product is outside specification limits. The number of defects depends on where th…[Read more]
Tony Burns replied to the topic Temperature/humidity effect on employees performance in the forum General 18 years, 2 months ago
I recall a very similar project, looking at the effects of lighting on productivity. Lighting was improved, productivity went up. Lighting levels were increased again, productivity went up further. Again lighting was improved and again productivity went up. Out of curiosity, the lighting levels were decreased – productivity went up again…[Read more]
You are on the right track, looking at controlling variation.
A problem with your suggestion is “then define how much I will allow the process to deviate”. You can specify a specification limit but the process will do it’s own thing. If the specification limit lies within control limits, you will produce defects. If your process is not…[Read more]
1. We do NOT “set” control limits as you suggest ! Control limits are calculated from process data. Control charts are “The Voice of the Process”. They tell us what is happening in our processes.
The only setting of limits that can be done is with specification limits. Specification limits are “The Voice of the Customer”. They are w…[Read more]
If we put aside the controversial origins of the +/- 1.5 sigma shift in the mean, any drift in the mean will cause points to fall outside control limits. A shift of 1.5 sigma will cause about 7% of points (depending on the shape of the distribution) to fall outside a control limit. Not something that would please customers! If every process has a…[Read more]
If the mean of a continuous process starts to drift, points will start to fall outside control limits. The process will no longer be “in-control”. It will become unpredictable and incapable of producing product that is within specification. If a process mean drifts to +/- 1.5 sigma, the process is well “out of control”. That is why contr…[Read more]
Yes, we have plenty of data, as well as more importantly, examples of the outcomes in business from application of the learning. We can also arrange for you to speak to some of our clients if you are planning in going down the e-learning path.
I think it is also important to appreciate that no one approach is a panacea. In…[Read more]
Prior to six sigma, PDCA (Plan, Do, Check, Act) was the core quality implementation methodology, however almost every major company had their own variants of this. A good thing about six sigma is that most companies have now standardised on the DMAIC variant. I suspect that all these date way back to the “scientific method”, what I might ca…[Read more]
As Michael says, the benefits of e-learning have been well documented: reduced training time and costs; more consistency; people learn at their own pace; easy management; no travel costs: train anywhere, any time.
However, Michael’s comment that “some people may get bored” cannot be underestimated. This is e-learning’s single greatest problem. I…[Read more]
Neither is correct.
Individuals chart (median moving range):
UCL = Mean + 3.145 * R tilda
Individuals chart (average moving range – more common):
UCL = Mean + 2.66 * R bar
The “3” in your formula probably comes from 3 * R bar/d2, or perhaps “3 sigma” control limits, upon which control charts are based.
It’s interesting to note tha…[Read more]
The requirement you have described in your latest post “…The SLA they want to set is 80% of calls answered in 90 seconds” is different to your original post: “80% of my calls in an average of 90 seconds”. This post implies no calls longer than 90 seconds, whereas that is not the case with an average, as described in your first po…[Read more]
Firstly forget about normal distributions. Time to answer calls is almost certainly going to be a quite assymetric distribution, with a long tail on one side and steep on the other. Fortunately this does not matter for control charts.
Where does your requirement for 80% of calls in an average of 90 seconds come from? Is this your p…[Read more]