Variation and Disorder

From the systems thinking paper we discussed the need to think of organisations in terms of a whole series of interrelating systems. A further complication is the existence of variation. Everything varies.

People have different characteristics, they vary day to day, they vary in relation to other people and in relation to different teams and different situations and finally they vary in relation to the organisation itself.

And then of course all the circumstances vary - markets vary, suppliers vary, the raw material varies, the plans vary, the equipment to make the product varies, productivity varies, the inspection process varies, the value of the pound varies, customers expectations vary etc etc

The need is to understand and limit this variation. The method used to appreciate variation is the "Control Chart" first evolved by Walter Shewhart in the 1930s.

In simple terms we plot the data on a graph as shown above so that we have a pictorial representation of variation. We then wish to define the variation present, so we use such terms as mean and upper control limit (UCL) and lower control limits (LCL).

To determine the control limits we start by calculating the mean or average. We then determine the average range which is the average of the differences between succeeding events/values/data (we ignore positive and negative differences i.e. ranges of -5, +4, +3, -6 and +2 have an average range of 20/5 = 4). And then we calculate what we call the Standard Deviation. A standard deviation is calculated by dividing the average range by the constant 1.1284.

From experience we know that:

We tend to use plus or minus 3 standard deviations to define the variation within a set of data. The Upper Control Limit (UCL) is the mean plus 3 standard deviations and the Lower Control Limit (LCL) is the mean minus 3 standard deviations.

In the chart we have indicated the results from a successful initiative to improve the design of the system. The second part of the chart shows greatly reduced variation.

We also differentiate between what we call Common Cause Variation (or noise) and Special Cause Variation (or signal). The Common Cause Variation (or noise) is the variation that is inherent in the system. The variation that is evident when the system is working as designed. Special Cause (or signal) is when unexpected variation occurs due to some an unforeseen circumstance - it signals an unexpected occurrence. We notice special cause when it appears out with the UCL and LCL. (The value of 37 in the above graph is outwith the UCL and would therefore be assumed to be special) The important factor here is that management action is very different when faced with Common Cause (or noise) as when faced with Special Cause (or signal). With Special Cause (or signal) we are dealing with a single unexpected result. We investigate that one event. We address Common Cause (or noise) when we want to improve the whole system - through redesign of the system.

It is a serious but all too common management mistake to take action without appreciating the difference. An employee is often criticised or held accountable for variations that are inherent in the system, something over which they have no control. It creates defensive strategies with a significant reduction in morale. The official term for taking action without understanding variation or the difference between common and special cause variation is called "tampering."

The above is obviously a very simple explanation of statistical process control (SPC). There is extensive literature available on the subject. However our recommendation is to:

  1. appreciate the concept,
  2. then design your systems (flowchart them)
  3. then collect data relative to these systems,
  4. use a simple graph to portray the data
  5. calculate Upper and Lower Control Limits
  6. Differentiate between common and special cause occurrences
  7. and take appropriate action to eliminate special causes
  8. work to improve the system to reduce common cause variation

In other words do the easy stuff first. When you have got to that stage then you can be more sophisticated in your analysis of data. Bearing in mind the objective is to gain knowledge from the data - especially in context of improving the system.

Please also bear in mind that we cannot eliminate variation. We can reduce it to acceptable proportions. We can reduce it so that we have a stable system.

In this diagram we are depicting two types of improvement. The first is achieved by systematically improving and tightening up the inputs to the system. We can improve training, machine maintenance, environment, clarity of instructions, quality of materials etc etc. But at some stage the improvements will flatten out. We will have stability relative to that design of the system. We may find this level of variation cost effective and acceptable. If not we have to consider a step change in how the work is done - we have to redesign the system.
Might we also portray the little information that is available when tabulated data gives us this month's figure compared with last month and the same month last year. It gives us no feel for the variation and capability of the system. It makes it even worse when we enter a target assuming that pressurising the individual is going to change the system.

If we set a target in the belief that it will encourage diligence then the individual has three choices:

If there is little acceptance of system thinking in the company and data is collected in tabulated form the individual is pushed into distorting the data or the system.

Another example of misinterpretation:
This chart shows variation with a mean of 22 and Upper and Lower Control limits of 31.75 and 12.25. Now if we want to encorage staff to score a higher score and think that people and their diligence is the main variable then we are tempted to set a target of say no less than 20, and if it does fall below 20 we 'kick arses' To the uninformed the policy works because improvement is usually evident immediately after a low point. We can kid ourselves on that our authority has made the difference, never mind the drop in morale and the bemused grumbling of the staff.

 

Disorder

Ontop of the complication of variation is the natural tendency towards increasing disorder. Consider a garden, we can expend considerable thought and work to create flowerbeds, lawns, herbaceous borders etc etc. But as soon we stop maintaining the garden, weeds and other plants soon invade the area and it quickly reverts to the wild disordered state. It is the same with our organisational systems, if the systems are not maintained then there will be a natural tendency for them to decline into disorder. In science we refer to this natural tendency towards disorder as The Second Law of Thermodynamics.

Let us consider this thought in terms of organisations. We have established that people come to work already motivated, wanting to be valued and take pride in their work. But they all have wills of their own and see things from their own perspective.

And the tendency is for each individual to beaver away in their own direction. We have all this energy but if it is going in contradictory directions then we have disorder.

In Thermodynamics they talk about heat engines that have a potential of say 1000 units of energy. They measure the efficiency of the engine by how much of that energy they can convert in useable power. They might be quite satisfied if they got 400 units of useable power from the potential of 1000 units. They refer to the units that they cannot recover as entropy. If their machines are not maintained then the units lost (entropy) starts to increase till the machine breaks down and the whole 1000 units are lost. The system reverts to chaos or disorder. Entropy increases to the full 1000 units

To use this analogy with organisations. Say the above diagram portrays a group of people that has the potential of 1000 units. How do we design systems that captures the maximum energy, and how do we maintain the system to avoid decay? And of course bear in mind that variation is all around causing confusion and chaos. The following table compares factors that decreases disorder and captures more of the available energy, and factors that increases disorder and loss. (entropy)

Decreases Disorder

(More energy recovered)

Increases Disorder

(Less energy recovered)

  • Long term thinking by the leadership
  • Development of a strong meaningful vision
  • Good communication
  • Good systems
  • Nesting of systems so that staff are aware where each system fits into the whole
  • Data that reflects variation
  • Involvement of staff in improving the systems
  • Good team working
  • Co-operation
  • Auditing to improve systems and the circumstances of individuals & teams
  • Flow mapped systems for ease of reading and understanding -reflects interdependencies
  • An atmosphere of trust
  • Etc
  • Crisis management
  • Compliance thinking (reduces energy willingly made available)
  • Top down communication, poor communication across the company
  • Poor systems
  • Unrelated systems where compliance is demanded without understanding the whole picture
  • Data designed for bosses - usually financial
  • Use of data to blame - produces defensive responses
  • Competition between individuals and teams
  • Conflict
  • Auditing to ensure individual compliance
  • Bureaucratic unreadable systems
  • An atmosphere of supervision
  • Etc

Conclusion

The basic assumption (Theory) is that variation exists and is a major influence in our systems. Not only does it exist but also there is a natural tendency for our systems to decay - to revert to disorder. We seek variety to move forward but need to limit and manage chaotic variation to capture the full potential of people/systems. It may sound paradoxical but when we reduce chaotic variation and approach stability the better prepared we are to address change.

home