Exvisu’s approach to the research we do is based on the use of actor network theory mixed with sound business research methodologies.

When analyzing any network, the importance of elements within it depends not only on the number of times an element appears, but on the degree of its connectivity within the network. By calculating the number of co-occurrences of specific keywords in a textual data set, we uncover the semantic links (i.e., we don’t rely on simple URL links) between actors and graphically represent the results to facilitate analysis.

The first step in Exvisu’s research process is similar to most consulting firms: we consult very closely with our clients to make sure we have a detailed understanding of the issues about which they need to learn.

The second step in our process is where we start to distinguish ourselves. Based on our consultations with clients, we build our own custom data sets that form the basis of our subsequent research. There are many potential sources we use to build the data sets: the blogosphere (via Google Blog Search and others), traditional media databases, academic publication databases such as Medline or the Web of Science, and others. In addition, we can build data sets from almost any custom database that our clients might have, including customer support logs, audit trail data, etc.

Once we have built the data set to study, Exvisu does advanced statistical analyses to learn everything we can about the data. The output of these analyses is one of two kinds of graphical representations or maps of an entire data set (or a subset we define based on client needs).

In each map, there are three relevant types of information: concept/query nodes, content item nodes, and links representing relationships between the two kinds of node.

Source map

A Source map presents an entire data set as a function of the source of the information in each record of the data set. For instance, a data set might be built based on a series of keyword searches in the blogosphere. The source map will present this data set in such a way that the semantic relationships between all nodes in the data set are clearly visible. Nodes that are more significant in the data set will be presented more centrally in the map and the relative distance between each node will demonstrate the relevance of it to any other node in the map.

Author map

Author maps present the data from the perspective of the authors of the information wherever they appear in the data.

Lexical analysis map

A lexical analysis map is a representation of all of the words in a particular data set that shows the relative importance of each word in the data and organizes the data into clusters of tightly related words

Methodologies

There are several methodologies that we use to ensure that the maps we produce contain valid, actionable research.

A/B testing

One of Exvisu’s most important methodologies is to perform A/B tests by building data sets that control for certain variables and output maps that make the results of each available in such a way that they can be compared. There are several applications for such tests, such as positive/negative association testing, pre/post activity testing

Longitudinal analysis

Exvisu can build data sets to perform “snapshots” of a particular data set at a particular point in time, but we can also update the data set at regular intervals