All processes, human and non, whether we are aware of it or not, are affected by variation. If we want to be able to predict the outcome of our processes, i.e. manage them, we need to understand, measure and manage the variation in those processes. The alternative is to manage by the seat of our pants. To function in a healthy and successful way, organizations do not need heroes of firefighters, they need managers that understand how to manage variation. Less glamorous, much more sustainable.
Intelligent Management and Statistical Process Control (SPC)
Let’s start by getting one thing straight: SPC is NOT a technique; it is a way of thinking, a mindset. SPC is foundational for Intelligent Management.
All processes, human and non, whether we are aware of it or not, are affected by variation. If we want to be able to predict the outcome of our processes, i.e. manage them, we need to understand, measure and manage the variation in those processes. The alternative is to manage by the seat of our pants. To function in a healthy and successful way, organizations do not need heroes of firefighters, they need managers that understand how to manage variation. Less glamorous, much more sustainable.
Intelligent emotions
The tool used in Statistical Process Control to measure variation in a process is commonly known as a control chart or process behaviour chart. These charts provide insight into processes and trigger rational thinking. After many years of relentless application of Shewhart’s charts, we can safely say that they are a powerful tool to manage intelligently the emotions triggered by the analysis of a process. We would discourage anybody from managing organizations using probability theory; what we suggest here is to leverage some basic, well proven, scientific ideas concerning the intrinsic variation of any human and organizational activity to take decisions that have an economical value.
In control and out of control
The way we deal with variation in a system is totally different in the case of ‘in control’ and ‘out of control’.
When a process is in control, it is working with the minimum variation possible given specific conditions of use. If these conditions remain stable, the process is in the best possible state as its behaviour is predictable over time.
In order to further improve this process we can only try to reduce its variation with the following actions:
- Stratify the data by dividing them into categories based on different factors and analyze how the data fall into subgroups
- Separate the data by dividing them into various categories and treating them separately from the others
- Gain experience by applying the Deming Continuous Improvement Cycle (PDSA): plan, do the experiment, monitor its results, learn from the effects observed and act
When a process is out of control, there is not a lot we can say about it.
Indeed, its behaviour is not predictable over time. It is subject to unpredictable jumps and all the data relating to it lose their predictive potential and become ‘historical’ data.
What to do when things are out of control
In order to act on an out of control process, i.e. try to bring it into control we must:
- Gather data as quickly as possible to identify rapidly the special causes that generate instability in the system
- Activate an emergency solution to limit damage
- Find out what made the special cause occur
- Implement a long-term solution
There are at least two excellent reasons for not wanting a process to be out of control. The first is connected to the impossibility of predicting which often makes it impossible to plan and carry out programs.
The second is linked to the costs associated with activities in a company that confuses common causes with special causes of variation.
Indeed, performance that seems good will often disguise poorly optimized use of resources.
However, this does not mean that ‘out of control’ is always bad. When you have a stable process, let’s say a sales data series which predictably oscillate within the upper and lower control limit, and we are trying to enforce a plan to increase sales, we actually hope to see the system going OUT of control on the upper side, so as to detect that an ACTION caused a shift in the system toward the desired direction, namely an increase in sales. By the same token, a process in statistical control is not necessarily a desirable process; oscillation limits that are too wide are often the result of poor understanding and execution of the process and force unnecessary costs on the system.
Again, the analysis must be performed with intelligence and common sense, and the charts have always to be read considering the operational context.
For further information on Intelligent Management’s approach to systemic management see our website www.intelligentmanagement.ws and our latest book ‘Sechel: Logic, Language and Tools to Manage Any Organization as a Network.
Leave a Reply