Implementation and Monitoring

Abstract

One question which should have been addressed during project planning is the mechanism that is going to be used to calculate model scores and apply decision rules based on the scores people receive. If a model has been constructed as a “proof of concept” or as a standalone research project then this may be a trivial question because nothing needs to be done. A second scenario is when a model is used to generate scores within an off-line environment. Within a target marketing context, for example, a model may have been developed to predict the likelihood of people responding to a marketing communication. The model will be used, perhaps just once, to produce a list of prospects to target. The list is then handed to the organization’s marketing team to upload it to a telephone dialer, e-mail generator or some other contact system. In this type of scenario the model, while important, is not “business critical”. If the production of the list is delayed by a few days, or the model is not quite as good as it could be, then the impact on the business will be relatively small.