Visualization for Large Collections of Ideas
Motivation
The open design of innovation processes (as described in the Open Innovation paradigm) enables companies to increase their innovative potential by utilizing external resources as well as concentrating their own development resources on the actual needs of their customers. By integrating a large number of external people into the innovation process, e.g. in form of online communities, a vast number of ideas for new products, services as well as improvements can be generated and collected in a relatively fast and cost-efficient way. Obviously, generating ideas easily is a big advantage of open innovation processes, but it is also a huge challenge with regard to the data illustration (visualization) and the data analysis. That is because usual forms of illustration like lists are not suitable for visualizing large collections of ideas, since they are primarily made for structuring data in a very simple way. Theoretically, a list may be able to handle an arbitrary number of (list) elements, but practically lists have only a limited potential in terms of visualizing data.
The Project
As an interdisciplinary student project at the chair of business informatics at Technische Universität München (TUM), we analyzed large collections of ideas with the objective to find suitable visualization methods. The work supports the presumed notion (that lists are not suitable) in its current state analysis and presents a catalogue of requirements (based on the analysis and the domain-specific problem) in order to design an ideal visualization for large collections of ideas. Requirements are identified by literature-based research and by interviews with experts of the specified target group. Subsequently, the visualization is designed in compliance with a well-known design and validation model by Munzner (2009) and implements the identified requirements as a prototype. The visualized data for the prototype is generated randomly and is pseudo-clustered (by topics) in order to demonstrate the different visualization methods (e.g. aggregation, sorting, filtering, coloring, clustering, and so on) in general. As a result, the prototype allows a flexible and interactive visualization even for large collection of ideas and features multiple views for the data like tables, a scatter plot, a timeline, and a graph. By this means, ideas can be visualized and managed in a suitable way compared to usual forms of illustration like lists. The visualization is implemented using the Flare Data Visualization library.
Screenshots
 |
| Table view: The visualization as table(s) shows two data grids. The first one is user-oriented and presents aggregated data (depending on the user), the second one is idea-oriented and shows all ideas. In both data grid, the columns can be sorted. In the idea-oriented table, it is also possible to select a particular idea in order to retrieve the content information for that idea. |
 |
| Detail view: Visualization of the content (id, title, user, creation date, ...) of a single idea. In general, users can get these information by hovering over a particular idea with the mouse. |
 |
| Graph view: This visualization type presents the ideas as a graph. By double-clicking on a node of a cluster, the user gets a more detailed visualization of the cluster (titles of the ideas are visible inside the visualization layout). Depending on the degree, the nodes (ideas) are colored categorically and the tooltip shows the maximum degree of the cluster. |
 |
| Interaction: It is possible to select ideas with the mouse in order so apply filters, e.g. remove selected ideas. |
 |
| Scatter Plot: The scatter plot is the most flexible visualization type. Ideas are presented as nodes in a diagram with two axes. Each axis can be configured separately and therefore allows the user to analyze the data. In the picture, the axes are set to "Number of Comments" (x-axis) and "Rating" (y-axis). Additionally, a linear color encoder is active which colors the nodes depending on their rating. Furthermore, a linear regression line is visible (that is possible since both axes are configured with numeric scales). |
 |
| Timeline view: The timeline view is specialized version of the scatter plot. It extends the scatter plot with user-based lines (edges) between nodes to create a timeline based on the user. In this view, the user can compare multiple timelines and trim timelines, i.e. filter/remove nodes of the timeline at the beginning or the end, by specifying the start and the end date in the popup shown at the bottom of the picture. |
* [Munzner 2009] Munzner, T.: A Nested Process Model for Visualization Design and Validation. In: IEEE Transactions on Visualization and Computer Graphics 15 (2009), no. 6, pp. 921-928.