mODELLING

FastStats Modelling

is the data modelling module, providing the user with the capability to predict the behaviour of customers and prospects.

When a customer profile or cluster analysis has been created, a model can be applied which scores, ranks and segments every record in your FastStats database. FastStats offers three main modelling techniques:

  • Profiling – using a patented Predictive Weight of Evidence (PWE) method that combines widely recognised Information Theory and Bayesian Probability, this technique scores individual customers and prospects and is fast, automatic and requires a minimum of user input.
  • Decision Tree Models (including CHAID) – This method produces a set of rules which are ranked to identify distinct segments or groups which contain proportionally more of your best customers and prospects. Decision Trees are particularly good for applying to external databases.
  • Clustering – Cluster analysis identifies groups of customers and prospects with similar characteristics. This method uses the K-Means technique to allocate each record to the nearest cluster centre, enabling you to better visualise and segment your database.

FastStats Discoverer already has a very powerful expression engine, which can be used to implement sophisticated models.  An example below where we implemented a propensity model we had initially developed in SPSS.  The power of FastStats is that this entire underlying code can be captured behind a single virtual variable, and made available across the platform to other users.