
photo credit: dsevilla
Adopting statistical thinking is critical for businesses, however, what is it?
Statistical thinking will one day be as necessary for efficient citizenship as the ability to read and write.
H G Wells (1866 – 1946)
When you can measure what you are speaking about and express it in numbers, you know something about it. When you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind.
Lord Kelvin (1824 – 1907)
Statistical thinking is based around three concepts:
The first concept is that all work occurs in a system of interconnected processes
The key words here are system and processes.
Here is a tool from six-sigma, called a SIPOC diagram.

iSixSigma SIPOC Diagram
- The outputs of one are the inputs of another
- The customers of one process are the suppliers to the next
Every product or service provided by company is the result of a combination or series of processes, whereby a variety of inputs come together to create one or more outputs. A process is therefore an added value transformation of input to output. It is the interlinking of basic processes which make up the complex systems of production.
The benefits of this are:
- It encourages focus on optimisation of the whole process, not just individual elements. Local optimisation could lead to imbalance or bottlenecks.
- It forces a customer focus – recognition of unnecessary complexity: which steps add little value (concept of waste in LEAN) - VSM
- It recognises the role of suppliers and the importance of up-stream prevention not down-stream detection (define terms)
- It promotes standardisation around best practice – again a lean concept
A couple of quotes from Deming summarise this concept nicely, he said that:
1/ If you can’t describe what you are doing as a process, you don’t know what you’re doing.
He also said that:
2/ We should work on the process, not the outcome of the processes.
By this he meant that the output from a process is dependent on the process itself and you need data to be able to successfully change the process.
Systems are several processes joined together. A system must have an aim and it must be managed. To manage the system, again, you need data.
However, processes will vary and because of this the output from any system will fluctuate. This leads onto the second concept of statistical thinking:
Each process in a system has variation in it. Like it or not, variation is everywhere:
- Journey time to work
- Quantity of products rejected
- Departure time of your plane
I’m sure you can think of several more examples?
All variation is caused. Variation is a natural phenomenon, it is everywhere, and because of this it can be quite difficult to say precisely what might be the cause of any particular example of variability.
All processes will be subject to variation in performance; no two outputs will ever be exactly the same. The overall variation is caused by variation within the process due to variation in people, materials, methods, equipment and environment. To regulate or improve a process we must understand the cause of variation. This does not mean that we need to measure the variation caused by every possible source; but we do need to know about the possible sources and understand which of them contributes most significantly to the output.
In any process there will be many (individually) small causes of variation which combine to produce a predictable degree of variation which remains reasonably constant over time, provided nothing arbitrarily changes the process (for example; traffic density affecting the journey time to work).
There may also be present significant “assignable” causes of variation. These are not usually a planned part of the process but which cause significant shift in process variation when they occur. Their occurrence can usually be “assigned” to something special occurring (for example, running out of petrol affecting the journey time to work).
All causes of variation therefore fall into two categories; those which are assignable (special cause variation) and those which are common to the system (common cause variation). Knowing this, helps in a very specific way because each requires a different reaction.
Any process may reveal both types of variation from time to time. For example, material quality will vary within a batch or from batch to batch from a single supplier. Hopefully, this will be controlled common cause variation which does not have a significant impact. However, variation in material quality from supplier to supplier may be much more significant and have a more marked effect on product quality. Changing supplier may therefore cause special cause variation. This decreases process reliability and predictability, and is one reason why “Single Source Supply” arrangements have increases in popularity over the last decade.
So, the third concept of statistical thinking is:
Understanding and reducing variation requires an understanding and appropriate use of statistical tools.
For example, statistical process control or run charts.
These help you eliminate special causes or unusual types of variation first then reduce common or expected variation.
They also help you to anticipate variation by designing robust products and processes in the first place.
The more certain we are about the way that people, equipment, methods and materials perform, the more accurately we can plan, manage and operate.
The idea of uncertainty about the performance of people, methods, materials and equipment might be little hard to grasp, it is perhaps better expressed as the predictability of processes.
If we are fairly certain about how a process performs, then we can predict its performance. Every process is subject to variation. No process is absolutely predictable, but the performance of any reasonably reliable process can be predicted within limits, provided nothing interferes arbitrarily with the process.
The first issue then, is the extent of these limits (i.e. the Process variation), because the narrower they are, the more predictable the process is. Reducing process variation increases profit.
Here are some simple equations which summarise what statistical thinking is all about are:
- Greater Process Variation = Greater Uncertainty = More Waste
- Lesser Process Variation = Greater Certainty = Less Waste
Which would you prefer?
Kelvin, process control, six sigma, statistical tools, statistics, thoughts