To provides a method and a device for estimating whether a specific event occurs or not after the occurrence of a trigger event to an exiting person.

In many cases, an estimate whether some event occurs or not and/or when it occurs after the occurrence of some chance (a trigger event) is required. If the estimate is made possible, an enterpriser can take action for preventing loss of customers. For example, customer data containing data relating to customers left from a bank after finishing of a loan is used to produce a Bayer' statistical model. In this model, plural attributes applicable to every customers exist, and these attributes are divided into plural segments. In one embodiment, a Bayes' statistical model is a survival analytical model, and in another embodiment, the model has a step for applying a Weibull distribution to data within each segment. A limiting likelihood of above data is calculated. This method mixes all possible segments inside a framework of a Beyes' method.