Volume 21, Number 1 Article by T Krishna Kumar March, 2009
Need for Cutting Edge Statistical Modelling for Business Analytics :
Businesses can benefit enormously from the statistical processing and analysis of data, such as in the assessment of risk and reduction in uncertainties in business forecasts, in targeting advertisements and sales, in designing production, in inventory and investment strategies, and in determining pricing strategies. Business analytics companies that specialise in data warehousing and macroeconomic modelling can provide firms with a competitive edge over their competitors. Yet, investment in statistical research by the business sector has been quite limited. Statistics is not used as widely as it should, in making business decisions. Compounding the position is the hiatus between the teaching of econometrics and the research in the field and their actual applicability in the business environment.
The routine approaches to statistical modelling do not meet the rigours of efficient modelling to deal with different patterns in different segments of data. The data generating process is insufficiently understood. Data is used as it is received, without being understood and processed. The theoretical scaffoldings of models in use are founded on a very limited methodological approach and better-performing alternative models are rarely explored. There is an ill-informed preference for 'objective' information or quantitative data.
Cutting-edge data mining methods with newly developed alternate modelling technologies are opening up new opportunities for statisticians in the field of business analytics. The underlying modelling approach that may be called 'robust modelling', and its superiority over the existing approach, however, has to be firmly established and widely communicated with a focus on relevance and scientific credibility which is the aim of this paper. Once the robust modelling with existing databases squeezes the maximum possible revenues, business firms will focus on gathering new data that is useful, and this in turn will create new areas of research in business analytics.
Reprint No 09106
