Last week we discussed using simple regression models for site performance models using our Demographic Dimensions database in order to reduce problems of multicollinearity. Often, the software that is used to ‘run the model’ are poorly integrated and may even require users to copy data from a site demographics report to the software that runs the model. Snapshot easily solves this problem with a powerful set of features.
Most models can be easily embedded into Snapshot variable lists so that the model predictions can be embedded into site reports and easily used to find hotspots. Simple regression models in the form Y = a1X1 + a2X2 + b can be directly entered as formulas within Snapshot templates using the REGRESSION() function.
Some models, however, are not quite so simple and require transformation of either the independent variables or additional computation before presentation. Snapshot offers a range of standard mathematical transformations that can be applied such as SIN(), LOG(), and SQRT(). Or we might be predicting sales per household and need to then multiply by households to get a sales estimate.
Other models take the form of a classification model using a sequence of rules, and in this case Snapshot can utilize a cascading set of TEST() functions to replicate the model logic.
The results can also be classified into target groups using the CLASS() function to allow users to more easily interpret the results.
Often there will be different models used based on the type of locational setting – sometimes even using different predictive variables. In these cases, we set context rules for trade area size (radius or drive time) based on density class, then can easily select which model we wish to output based on the site type. With just an address or coordinates, it is possible for an application to create a report that determines the appropriate trade area size (based on density class), selects the model to be run, and classifies the site potential, and presents the results without clutter.
By putting all of these elements together, Snapshot can be easily used to map potential site performance in a market to create powerful integrated site analytics tools.
Recent Comments