- Statistical Process Control or SPC is sometimes taught containing quite a lot of maths. But actually, the principles behind it are simple and very useful to understand. So, let's have a look at them without any maths. First, the idea of Process Control. If you inspect things after you've made them or if you're providing a service, the inspection would be after you've finished the work, like how well a room is cleaned or how well a meal is cooked, then it's already too late.
I suppose you can learn for next time and do things better as a result of the inspection, but it's not really as good as controlling the processes. Knowing that your oven is the right temperature and the turkey is in there for the right amount of time is much better than going, "Oh dear, the turkey is burned." So, the idea is that we know the temperature and the time and we make sure we get that right every time. We might even experiment with a hotter oven for a shorter time, a cooler oven for a longer time, til we find out how to cook the perfect turkey and then stick to that every time.
This is process control and it's easy for a turkey, but it's tricky for cleaning or writing software or making a training video. Somehow, you have to find out what the key ingredients of the process are and then, make sure you get those right. That's where the statistical part comes in. Have a look at this target. Two marksmen are firing their rifles at the target. First, marksman A has shot as follows. He has hit the target every time and his the bulls eye once.
Then, marksman B has shot as follows. She has missed the target once and hasn't hit the bulls eye at all. But who is the better marksman? Yes, it's marksman B, because her grouping is better. Grouping is a term for how much variation there is between her shots. The fact that her grouping is already good means that her results are quite easy to improve. She just needs to adjust her sights and she'll be spot on. Marksman A on the other hand, oh dear.
To improve him, he needs a new rifle or maybe a new eye or a new, more steady arm, which isn't going to be easy. So, basically, there are two reasons for failure, which are drifting of the average, that's the sights out of adjustment, or too much variation. That's the unsteady hand. To just inspect the target and say, "You missed," doesn't tell us why. So, we don't know what to change in order to get better. If we did everything by the recipe but our turkey is burned, it means that either the oven is out of calibration or the temperature is fluctuating in the oven.
But which? To find out whether it's drift or variation, we need to do the equivalent of assessing the grouping. So, we take four random readings and see how spread out they are. If they're all very close, we know it's the calibration of the oven temperature. But if they're all over the place, we know the temperature control is failing. Also, this sample of four gives us a measure of what the average is without being misled by the occasional blip.
And if we take samples every hour or every day, we can track whether the average is drifting. So, the sample of four allows us to track both drift and variation and we then put both of these on graphs so that we can see what the process is doing over time. And even better, not only does this tell us why we failed and what to do to fix it, but it also gives us warning of problems before we fail. If our oven is starting to drift out of calibration, the SPC graphs will tell us long before we can taste anything wrong with the turkey.
So, we'll never get to the embarrassing point where the Christmas dinner is a disaster or something is sent to a customer that's wrong. Imagine if you tracked the time taken to serve customers or the number of typo's in reports. You'd know if things were getting worse hopefully before there was a complaint. The key is that it has to be measureable in order to apply these statistics and draw the graphs. So, if you were say, an architectural practice, you want to get an actual number for how much the client's like the designs done by each of your architects and then track them on graphs so you could look for the average.
Is it drifting? And also, how much spread is there? How much does each architect vary? Are any getting worse or more variable? Do you want one that averages high but is controversial or one that averages only medium, but is consistent? At least you'd have the facts in order to be able to decide these things. I think you can see that pass fail doesn't come close because it's too late and it gives you no real understanding of what's actually happening in your process.
So, that's SPC and of course, the question is, how can you apply it to your part of the business? What are you key processes and what are the indicators within those processes that you could measure and that you could track on graphs over time? These could be things like costs, waste, hours taken, overall deliver lead time, percentage delivered on time, quality accuracy or customer satisfaction. It sounds like a bit of work, but please do try SPC.
It's one of the best and most neglected process improvement techniques.
Lynda.com is a PMI Registered Education Provider. This course qualifies for professional development units (PDUs). To view the activity and PDU details for this course, click here.
The PMI Registered Education Provider logo is a registered mark of the Project Management Institute, Inc.
- Recall how to measure processes.
- Discover how to use statistical process control.
- Assess the quality, cost, and time trade-off.
- Analyze methods of reducing cost by reducing waste.
- Identify how to improve delivery time.
- Define the Lean and Six Sigma processes.