Previous article: How Often? – BI General Question (Part 4)
This is the final part of a general introduction to BI using a case study of a mobile phone store.
How Likely? – This group of questions require more complex analysis of past data to identify trends in order to predict future events. This can be done on sales figures, market share, or various other parameters that affects the company’s performance. This analysis can be presented in a simple tabular form, or even as a graph that reveals the past present and future trend. There is a fine correlation between the accuracy of the data and the prediction, however, a good analysis requires a weighted contribution of historical data. The older the data the lower its contribution to the prediction of a future event and inversely, recent trends and events have a larger impact on the prediction model. This kind of analysis is called predictive modelling.
How Quickly? – This kind of question is again based on more complex analysis and is also related to the ‘How likely’ scenarios from the previous question. Finding out how quickly certain milestones are going to be reached requires an underlying understanding of the likely consumption rates of the particular units being analysed. The answers to these questions impact the financial models of a company and are used to plan the investment opportunities that can be reached out for. This analysis also falls under the predictive modelling category.
There are many predictive analysis tools and methods used for this kind work, however, these methods are only as good as the understanding of the analyst in what they are doing. There are many predictive tools, even in a simple excel sheet the modelling equations offered on a graphical representation of a set of data is quite advanced, however if these are used with little understanding of the statistical errors built into them, then the understanding of their graphical representation can be very flawed, leaving the door open to a lot of misinterpretations. Hence one should always proceed with caution when studying the statistical results of an analysis.