Descriptive, predictive and prescriptive analytics – An overview
“More than the past, it is the future that interests me, since that’s where I intended to live” is a quote from Albert Einstein which in connection with the winged in German saying “knowledge is power” increasingly gained in importance in industry and trade. So corporations try to collect, in the context of the Big Data movement, as much data as possible to be able to make predictions about developments in the future. This is attempted with various analytical methods from the field of Business Analytics (BA). The term BA is described at Wikipedia as follows:
“Business analytics focuses on developing new insights and understanding of business performance based on data and statistical methods. In contrast, business intelligence traditionally focuses on using a consistent set of metrics to both measure past performance and guide business planning, which is also based on data and statistical methods.”
Among the most important tools of BA are:
Descriptive Analytics is designed to collect historical data and to answer the simple question of what happened in the past. In essence, this approach is based on historical events, by analyzing the factors which made projects successful or not. So it is based on learning effects. The major part of data, approximately between 80 and 90 per cent of the data used today, are based on that system. An example is that a library recognizes that it has lost customers in the last three years.
Predictive Analytics, however, go one step further than descriptive. The goal here is not only to explain what has happened in the past, but also to draw conclusions from the available data. This is done by the use of numerous statistical methods, such as predictive modeling, machine learning, and data mining, which analyze old and up-to-date data to make predictions based on historical contexts. Here is an example that a library mainly loses young customers due to a lack of children’s and youth books.
Prescriptive Analytics is an enhancement of the predictive analysis method. In this analysis model, on the basis of elaborate analytical models and Monte Carlo simulations, which, in addition to the future prognosis, also give concrete recommendations for action to avoid or occur an event. In this practice, known and random variables are evaluated. As an example, here the finding of the absence of young customer due to an insufficient offer of children’s and youth books, paired with the recommendation for action to acquire further and perhaps to set up a separate reading for children and teenagers.
In particular, the methods of Predictive and Prescriptive Analytics are becoming increasingly important in industry. While Predictive Analytics is already widely used in commerce, insurance and finance, the manufacturing industry is starting to deal with its advantages. It is hoped to reduce storage costs, optimize the maintenance of machines, to increase the product quality and availability of goods. As an example, it is conceivable for an automobile manufacturer to carry out the material requirements planning on the basis of the selected combination of components at the online configurator as well as other internal and external data (e.g. the season). So it is possible, that because of the indicator summer, combined with the frequent selection of a specific aluminum rim in the configurator, the analysis program reports an increased demand for this rim. In this case, prescriptive analytics would serve as a concrete recommendation for increasing purchasing volumes of this rim, while at the same time reducing others. The result of this advice is the increased availability of rims, while at the same time lower costs (e.g. for storage).
In conclusion, therefore, the Descriptive Analysis is a summary of data in order to assess past events, while the Predictive provides concrete statements for the future. Prescriptive Analysis can be regarded as a further development of the predictive approach, since it also provides concrete recommendations for the prediction. However, it can be concluded that there is no tool which is capable of predicting the future with a 100% certainty or even giving recommendations for action with a corresponding correctness.
Ultimately, these are tools giving recommendations for action. The final decision, what is done with this recommendation, is left to the man, who has to be responsible for the consequences.