Behavioural modelling - key terminology

BASE GROUP - people you wish to compare the analysis group against.
ANALYSIS GROUP - a subset of the base group, identifying people who have already demonstrated the desired behaviour.
EVENT - a step along the customer journey, identified by the occurrence a particular transaction linked to the customer.
EVENT DRIVEN DATE - a date which can be unique to each individual, identifying the date an event occurred for them.
TRAINING DATE - the date around which yo study records in order to create a model.
SCORING DATE - the date when you apply the model. Once you have found a model you are satisfied with, this date is the date on which you create the virtual variable and move into 'production'. Note that you can also score data in order to evaluate it.
REFERENCE DATE - the date from which all the eligible transactions are identified in the selection dialog, to inform which people are included in the analysis and base groups. It can be any of the training, evaluation or scoring dates depending on the stage of the modelling process.
POINT IN TIME - this date is the starting point from which behaviour is analysed and can be different for the analysis and base groups. For example, the base group might use Booking Date, whilst the analysis group might use Policy Date, meaning you would study the behaviours prior to making a booking in the base group, but prior to taking out a policy in the analysis group. This date is normally the event driven date.
BEHAVIOURAL FEATURE - also referred to as a 'dimension' - a FastStats expression, generated on the Dimensions tab, that typically explores transactional data.
RECENCY - e.g. Time since last...
COUNT - e.g. The count of...
VALUE - e.g. The cost of,,,the change in mean...
CRITERIA - specific criteria that you add to count, recency or value dimensions. For example - the count of bookings made by people who have booked certain facilities.
COMBINATION - a specific combination of transactions.
PWE (Predictive Weight of Evidence) - a measure of how predictive of behaviour each category within a dimension is.
MEAN PWE - the average predictive strength across all categories in a dimension.
INSIGHT PWE - differs from Mean PWE because Mean PWE is across all categories. To calculate Insight PWE, categories offering no insight are removed. Then the PWE for those categories offering sizeable and significant insight is calculated, weighted by the number of people, and an average taken.
POWER - a figure on the Results tab, ranging from 0-1, used to determine which model is the most predictive. The higher the number, the more predictive power the model has.
ASSOCIATIONS - how similar are the dimensions? Associations allow you to determine if two dimensions are so alike that they could potentially bias the results.
NICHE FEATURES - features which apply to only a few customers, but still provide a strong indication of being a good prospect, or not.
COVERAGE (Dimensions) - the number of base records included in a dimension.
INSIGHT COVERAGE - provides an indication of the number of people for whom a dimension can give a significant and sizeable prediction.
INSIGHT TYPE - all features are assigned an insight type based on the nature of the insight they provide.

INSIGHT COLUMNS - you can view the above concepts and measures as columns within the Dimensions panel.
INCREMENTAL INSIGHT - a single metric for choosing dimensions that are predictive and diverse. Predictive distinguishes good / bad prospects from average, whilst diverse finds multiple unrelated ways of identifying prospects. When dimensions are ordered from highest to lowest by Insight PWE, all dimensions can be considered together to provide incremental insight. The dimension with the maximum PWE maintains the same value, but all other dimensions are adjusted and their values reduced as any insight shared with variables higher in the list is removed.