Spotfire's DecisionSite aims to accelerate and sharpen decision making associated with product design, development, and manufacturing, by facilitating analysis and pattern discovery through visualization. With the goal of creating a platform that can be "verticalized" to support a wide variety of analytic domains, Spotfire has incorporated an extensive feature set into the base product. It managed to do so within a user interface that's both innovative and effective.

For this review, we worked with the generic DecisionSite product, and took brief tours of versions customized for life sciences applications, which arrive with predefined, domain-specific analytic templates.

DecisionSite operates against a tabular data set, which you typically acquire by importing data. DecisionSite can import from local text, spreadsheet files, or external SQL databases, and converts the imported data into its proprietary format. The product ships with more than a dozen statistical tools (such as hierarchical clustering, decision tree, and profile Analysis of Variance [ANOVA]).

Creative Visualization

The core of DecisionSite is a set of visualization tools working with an engine that creates a set of graphical data filtering controls, which Spotfire refers to as Query Devices. (See Figure 1.) Among the visualization tools are several charts and a tabular data view, which have display characteristics that are highly configurable through a multitabbed property dialog. As we describe its features, imagine each one as being easily integrated with the previous feature to create an effective and playful environment for performing real-time visual drill-down and pattern exploration.

When it first loads a data set, DecisionSite creates an initial graphical presentation based on a series of rules that consider the number of data points, the data types of the columns in the data set, and the number of different values in columns. The available visualizations include 2D and 3D scatter plots; bar, pie, line, and profile charts; and heat maps, histograms, and tables. You can use the trellising feature to split any of the visualizations into panes. Each pane contains a subrange of a single attribute (such as gender or age group).

A profile chart maps each record represented as a line, with each point representing the value of some attribute. Profile charts can be useful for comparing the degree of similarity among a number of records. For instance, each line might represent a customer, with each point along the axis representing a discrete behavioral measurement. Or a line could represent a pharmacological compound, with each point along the axis measuring some chemical characteristic. Also, you can use profile charts to present comparisons of the states of various items over time.

Heat maps, also known as intensity or matrix plots, present a tabular view of a collection of data points, with different colors representing specific data values or ranges of values. Heat maps allow you to quickly identify clusters of records with similar values, which are displayed as "areas" of similar color. You select the columns to include on the Heat Map Columns tab in the Properties dialog and customize the coloring on the Colors tab. (See Figure 2.)

Limiting Options

DecisionSite uses the term "Query Device" to refer to the visual controls that allow analysts to restrict, or filter, a data set. There are six such Query Devices. A radio button device represents Boolean values or attributes having a limited set of discrete text or numeric values. Checkbox devices represent attributes that take a limited number of discrete text values. An item slider device represents attributes that take a larger number of text values. A range slider represents a large number of numeric data values. A full-text search control provides a way to filter text values based on your criteria.