Stimulation Initiative for European Neural Applications
Esprit Project 9811
Modelling Market Dynamics in Food-, Durables- and Financial Markets
Company background
BrandmarC B.V. is a local Dutch consulting company. It's core and sole
business is researching and developing precise and reliable analysetools
for nonlinear problem domains, especially the area of marketing and
communication. About 50% of the top 20 spenders in marketing and communication
belong to the clientele of BrandmarC B.V. .
The problem
Most markets behave nonlinear. That is an advantage for marketeers,
but also a threat. The main concern is directed towards the uncertainty about
the effectiveness and efficiency of marketing- and communication-efforts
under different circumstances. Context plays an important role with regard
to the importance and relevance of instruments (variables). With the
techniques most commonly used in marketing environments it is not possible to deal with this context problem. Therefore, research and
analysis in marketing and communication is not supplying the contextually
information needed as input for the decision making process.
A hybrid application using neural networks
BrandmarC developed a general software tool dedicated to the problem described
above. A lot of effort is dedicated to selection techniques to reduce data to essential information, before the actual training of the neural net (model)
takes place. Different tests show that selection techniques are crucial
with regard to reliable outcomes of the final model. Other tests
prove that neural nets are far more precise and reliable than
statistics in predicting in the field of marketing and communication.
The IDL induction algorithm, programmed by BrandmarC in cooperation with
the A.I. lab of VU Brussels, is in most cases very consistent with the
neural net model. Up till now reality proves that the results of different
models are correct.
Benefits
The gap between the information gathered by linear statistics and the
information gathered by neural nets is so wide, that we assume that in
many cases the use of linear statistics leads to wrong decisions.
Next to that, clients have a better insight in the (ir)relevance of data,
the relevance of instruments in diferent contexts and are equipped
with scenario analysis to enable a quicker and more efficient decision making
process.
Generalization
It proves to be possible to analyse data from different brands with different
positions in the market place in one model. The neural net learns contextually
and is able to discover unique positions of brands as well as generalisations
stemming from all brands in one model. These findings are seen by BrandmarC
as a breakthrough in nonlinear research. The uncertainty in marketing and
communication with regard to the relevance and importance of instruments
in different contexts can at least partially be solved.
Contact person
R.J. Schuring, BrandmarC B.V., Amsterdam, The Netherlands