Abstract

The early bird catches the worm, but is caught by the early cat.

The behaviour of complex systems is, of necessity, difficult to predict. Very often our analysis of a potential change is deliberately limited to a particular part of the system. A serious consequence of this is that changes may fail to produce the desired results when this outcome was clearly inevitable. Changes that were intended to produce some result often turn out to be totally counter-productive. The way to avoid this type of error is to ensure a 'Total Effect Analysis' is performed in addition to any other more specific evaluation.

1.   Terms

1a.   Complex System

A 'complex system' is defined here as any system which can be divided into sub-systems, each of which has a degree of autonomy and interacts with other sub-systems.

Common examples of such systems are:

• businesses which operate several profit centres;
• research laboratories;
• schools;
• countries; and
• any organisation with several managers

Note that the activities of the different sub-systems affect each other. This is, in general, a realistic assumption: the sub-systems will be competing with each other for any resources allocated centrally, and they will each be operating in the same world. If there is no central allocation of resources and no effect on each other there is effectively no larger system, just a collection of independent smaller systems.

1b.   Sub-system

A sub-system is itself a system, and can be a complex system. This breakdown can go as far as is helpful: A company can be formed of several cost-centres; a cost-centre can be formed of several departments; a department can employ several managers; a manager can supervise several employees; an employee can be responsible for several projects; and so on.

The question is not whether a system can be broken down into sub-systems, but whether it is helpful to do so in the current investigation.

When terms such as 'other sub-system' are used here, they should be understood to mean a 'sibling system' - not just any sub-system, but a sub-system which is a part of the same larger system.

1c.   Total Effect Analysis

'Total Effect Analysis' is an approach to investigating proposed changes which takes into account the interactions of two groups of sub-systems: firstly, the sub-systems within the system being considered; and, secondly, all the sibling systems of the system being considered.

2.   Total Effect Analysis

The basic question to ask is: apart from the factors controlled by the system being considered, what factors affect the output of the system?

2a.   Sub-optimisation

One aspect of Total Effect Analysis is the need to sub-optimise the behaviour of sub-systems within a system in order to optimise the behaviour of the system. This is in contrast to the generally accepted assumption that system optimisation can be achieved through optimisation of the component sub-systems.

Thus, a business may be considered (at a superficial level) to have a goal to generate maximum profits. These profits are all generated through the various profit centres. The business instructs each profit centre to generate maximum profits. Instructing the profit centres to optimise profits is a sub- optimal strategy for the business.

This is, of course, overly simplistic. A business has to generate profits in this financial year, while retaining customer loyalty in the next financial year, and while building an infrastructure that will enable even more profits to be generated over the next five years. But, however complex the goals of the business and however sophisticated the strategy to achieve those goals, if the business simply passes the goals on to each of the sectors, the total effect will be sub-optimal for the business as a whole.

The Two Salesmen

It is easy to illustrate this problem. Imagine a company which sells a single range of products, and employs two salesmen to sell the products. The salesmen work largely, or solely, on commission, so they each have an incentive to sell as much as possible - that is what the company intends. It is often assumed that this arrangement will maximise the company's profit, but is this really the case? To keep the analysis simple, we can ignore the unlikely possibility that there is precisely the right number of customers to keep two salesmen fully occupied.

If there are enough potential customers, the two salesmen can divide up the territory in any mutually convenient manner, and each gets on with selling as hard as they can. In this situation, the company is missing potential sales because not all the potential customers are being reached.

The other possibility is that there are not enough customers to keep both salesmen working as hard as they want. This time, the salesmen are competing with each other for customers. They will undercut each other in an attempt to get the business (and thus the commission), and therefore produce less profit for the company.

Of course, if the salesmen are competing, they may do various underhand things which would reduce the company's profits. But the point is that the organisation has set up a system which has the inevitable effect of motivating their employees to engage in behaviour which will harm it. Perhaps in real life the salesmen will be moral enough not to do anything which would attract sales away from the other person... perhaps. But in this case, the company benefits despite the system, not because of it.

The General Situation

This example of the two salesmen can be generalised to any organisational structure beyond the completely trivial.

It is commonly encountered as conflict between departments: the salesmen are rewarded for selling products which manufacturing have not yet made; manufacturing are rewarded for shipping products before they are ready, which causes a problem for customer support; and quality assurance are rewarded for identifying reasons for not shipping products to the customers at all.

More Sub-optimisation

The previous examples are all easy, in that they can be predicted if you just think about what you are asking people to do. Life is not always that simple, because you don't know everything about the system.

Numerous companies have tried to make their staff more productive by reducing the amount of time they spend waiting at the coffee machine or water cooler - by providing more drinks machines, arranging for drinks to be brought round on a regular basis, and so on.

But they have often found that departments where these changes are made become less productive. They found that employees used to share useful ideas or vital information while they were waiting to refill their coffee cups, and this passing on of information was much more important than the lost productivity.

There is also the possibility that changes to increase productivity may make the employees feel less valued, or may seem petty or bureaucratic, and may result in less motivated staff.

In each of these ways, seeking to maximise the productivity of the parts of the system may result in a decrease in the productivity of the system as a whole.

2b.   The Impact of Other Sub-Systems

Taking into consideration the principle of sub-optimisation, the first area of concern is the impact of my proposed change on the other sub-systems.

If I sell more, will we be able to support the extra customers? If I cut the price of my product line, will it make another company product line appear over-priced?

A change may make ny own department more profitable, but what will be the effect on the other departments?

2c.   The Impact of Feedback

Simple feedback is built into economic models, but this often does not go far enough.

For example, if I double the price of my product, I will not expect to double my income, as some people who would have bought at the previous price will not buy now. Of course, this is not a simple rule: sometimes when I double the price I get more sales because the public percieves the product to be worth more.

But, in either case, feedback is operating - double the price does not result in a simple doubling of the income. What I do affects my customers, and it affects the market as a whole.

Care in the Community is an obvious example. People with mental health problems were living in massive institutions, which cost a lot to run, and delivered a lower quality of life than supporting the patients to live in the community. But the systems to care for them were not put into place, so many of them ended up in prison, which costs far more than the institutions they left, and where the quality of life is much lower.

The basic question is: if I make this change, what other changes are likely to take place as a consequence?