Conventional Predictive Analytics: The Good and the Bad

Data preparation is required for conventional predictive analytics. On the basis of educated guess users are required to make predictions regarding the risk variable which requires to be analyzed. Many things are left unknown and only a subset of the interrelationships can be analyzed and this is due to limited time and cost.

Identifications of the contact strategies, most effective collection agencies can be done through predictive analytics which optimizes allocation of collection resources. In the context of forecasting and predictive modeling SAS technologies helps in improving operations. In addition to that decision-making, efficiency and productivity can be optimized.

When the customers has initiated the procedure for terminating the service, then only the business take actions by means of response to attrition of the customer. The chance of altering the decisions of the customers at such stage becomes difficult and impossible. Proactive retention strategy can be made by properly applying predictive analytics. Predictive model is capable of determining how likely a customer is going to stay or how likely he's going to leave in the future. And this is done by examining past service usage of the customer, spending behavior, service performance etc. Lucrative offers can lead to increasing customer retention. There is another problem that most companies face; this is when silent attrition takes place. By silent attrition occurs when the customer slowly yet steadily diminish the usage. For increasing customer activity predictive analytics assists the companies to take the right action, and this is possible because this analytics have the capability of predicting the future behavior.

Predictive analytics projects deals with statistical techniques, statistical modeling, proprietary models, complex manipulation, tools of data mining etc. There are a few numbers of experts with such skills and thus it becomes expensive to hire them. The fact is that person involved with the business data is intermediated with the statistical experts and is not so involved in processing. Predictive analytics projects require the following:

• Preprocessing


• Integration support


• Significant data preparation


• Internal data source management

For yielding an outcome or result conventional process may take different time periods for example weeks or months which is dependent on the state of company's data and the resources that are available. Statistical experts have to wait for data because sampling and extraction of data is required for model building. Deployment cost, accuracy of model, interpretability, and score latency are evaluated by the analyst.

Conventional Predictive Analytics - The Good:

In back testing on a sample predictive model is evaluated with the help of conventional approaches. And such measure leads to overstating the accuracy of the model that is achieved at the time of deployment. Also, significant business value can be delivered through conventional predictive analytics.

Conventional Predictive Analytics - The Bad:

In some cases, more damage than good happens when the target is made on the wrong customers. In pervasive business applications of predictive analytics there are two main demerits that are noteworthy.They are as follows:

• Targeting direct marketing


• Targeting retention efforts