Neural network example

Neural network example (Photo credit: Wikipedia)

When we usually talk about business intelligence, we talk about a set of tools that tries to acquire data, to process them and provide information useful to make better decisions about the future of the company. This is not what I am going to present in this text, I am going to talk about the behavior of a company as a live being. It is very risky to consider a business as a live being, but if we go back to the classical definition of it that a teacher gave us at the school, it seems more appropriate.

A live being is born, it grows, it is reproduced, and finally, it dies. In a similar way, a business company is born, sometimes it grows, sometimes it is reproduced in other countries or markets, and finally it can die.

I have not increased the knowledge in the world about businesses with this last sentence, but I want to go a little farther. I am not interested in a biological viewpoint of a company, only in biological viewpoint of its management. The former viewpoint can justify why it is important that a business is operating in a “healthy” state, but the latter can help us to understand better some aspects of business organization.

What we named structure in a business are a set of people that have the responsibility to acquire, to process and to transmit business information. We usually name managers to them. There are a lot of different focuses to analyze the competences of a manager, but following this biological approach two ones are needed. A manager must have the capability to communicate and to process information.

The first one is related to his ability to contact other people downstairs and upstairs, to understand their work, what is the really important and to simplify the data to be served to the upper management layer.

The second one is related to the analytic work. A manager needs to turn the operative information of his department into the proper information for upper level, in general in financial and strategic information.

This scheme reminds the way as the nervous system act in the human body, and we know that the knowledge of a human body is very different in quality and quantity to the knowledge of the sum of its neurons. We can affirm that this is true in a business company too.

In a neural network, knowledge is distributed all over the network. A company can work effectively although nobody at the company knows how all processes are working, however the full company can do the full work.

In a small enterprise the knowledge on the company usually is fully centralized in one or two people. This has the advantage that the decision making process is done for a natural human brain. When the data are huge as in a large company the knowledge and the decision making process is distributed among a set of managers that act as a single neuron instead a full brain. The advantage of this scheme is larger amount of data that can be processed.

I have followed a descriptive analysis so far, but the interest of this discussion would be to provide some useful knowledge to be used in a business.

Companies usually try to reduce the number of management layers. With the modern computer systems this fact has been possible because they have increased the processing capability of a single manager as if they had multiplied the number of connections of a single neuron.

With this biological approach we can analyze if this scheme is valid. I am going to introduce you the world of the artificial neural networks, these are computer programs that process information in a neuron-like scheme. The most known ANN is probably the multilayer perceptron. It has been demonstrated mathematically that a MLP with a single intermediate layer (hidden layer) can learn and store the same knowledge as another MLP with more intermediate layers.

In terms of complexity, a single intermediate layer is much better than a lot of ones if both provide the same work, because there is always an increase of uncertainty in the transmission of information from one layer to another and in every processing center.

We should trust in the simplification of structures through the reduction of managing levels, however, typically, a large enterprise has oversized the number of layers generally although this is totally inefficient. The reason is not an increase of the effectiveness; the reason is a lack of the proper human resources.

With a single layer structure, the communication with the lowest level requires operational knowledge, and the communication with the highest level requires financial and business strategy knowledge. However we can increase the structure of the company all we want with personnel that only process financial data, but if they do not provide value, they are only mouths to be fed.

The number of managers never is selected because of the managing needs (processing needs); it is selected, as in nature, by geographical needs (physical reasons of information transmission). As in the nature there are two kinds of neurons, nerves and brain, middle managers and directorate staff. But remember, if nature had developed radio communications perhaps it would not be necessary having a long nerve from the hand to the brain in order to move a finger.