CS 635 - Final Report
A Web-based Visualization Tool For Analyzing Multi-dimensional Data
Contents
HCI Systems, Inc. is a medium-sized electronics manufacturing firm building circuit boards for use in modems, home security and general electronics products. HCI uses a direct sales force to sell directly to manufacturers. Salesmen and managers both spend a great deal of time on the road working with customers. Because most manufacturers now use just-in-time inventory methods, there is always concern on the part of potential customers over the ability of HCI to deliver product in quantity and on time.
Current systems at HCI require more than a few minutes to retrieve needed information over 80 percent of the time. As such they are not suitable for answering customer questions and are not responsive to company management in general.
The Project Team was appointed to develop an information retrieval system that would be accessible from multiple locations, provide needed information quickly, and simplify complex data. An initial iteration prototype of such an interface consists of an ActiveX control imbedded in a web page that is accessible from the company's main web page.
A user profile was developed using a questionnaire distributed to a sample of potential users. The complete report is at Appendix C.
The typical user is either a professional or a member of management at some level and is college educated—with at least a bachelor's degree. He or she is comfortable using web browsers and expects to use the system in a variety of locations (home office, customer sites and on the road). The preferred way of viewing complex data is in the form of charts, tables and graphs.
The project team identified eight tasks that users would need to perform in conjunction with using the system:
Usability Objectives and Requirements
Criteria were developed to evaluate
The principle requirements were that a user would need to be able to access the interface from a customer's office and retrieve the necessary data before the customer loses interest—a matter of minutes.
Usability was measured in a two phase process, beginning with a subjective questionnaire presented to a focus group following a presentation and demonstration of the prototype followed by usability testing against criteria which measured ease of use by the ability of users to perform tasks with in a stated time constraint or number of mouse clicks.
The Data Visibility project is developed using the Star Life-Cycle Model. This approach contains many of the same stages of development as the better-known Waterfall model, but allows for continual evaluation and the revisiting of stages as required. This approach allowed the team to rapidly develop requirements, design and produce a prototype, since the requirements will be revisited in light of information gathered in the evaluation of the prototype.
Team decisions were reached on the basis of consensus and at no time did it become necessary to resort to voting. Major development decisions were reached by meeting in committee, conducting open and frank discussions, and reaching agreement.
At present the system is a demonstration quality prototype relying on built-in dummy data rather than live access to the corporate data warehouse. While this is adequate for preliminary evaluation of concept and user interface, final evaluation of the concept and interface will require use of live data and more freedom for the user-tester to experiment.
Testing indicates the product is usable but requires further refinement.
Survey: Although the favorable response rate was, in all cases, greater than 75 percent, a significant number of users had difficulty with the initial explanation, understanding the controls and options, or understanding the information presented. There were at least 10 recommendations relating to adding labels or instructions to the interface. Design decisions relating to the affordance of the controls will have to be revisited. In addition it is clear that the planned training video may be required early in the process to properly evaluate the understandability of controls, parameters and interface.
Usability Test.: At least 60% of test subjects, completed all four criteria based tests successfully. However, 2 of 5 had trouble choosing the appropriate set of data and selecting data dimensions. Three had trouble remembering to click ShowData. This result confirms the conclusions drawn from the survey data about the affordance of the interface: it is not as clear as we had thought. In addition further attention must be given to the design and placement of the Submit ('ShowData') button. This button seems to be required by the technology that we used, but alternatives should be explored. If the button cannot be eliminated, it should be made more prominent.
Often business managers are faced with the problem of analyzing large volumes of data and deriving knowledge that can be used to guide a business venture. Modern database management systems now provide tools such as Online Analytical Processing (OLAP) engines that help by summarizing data over various dimensions or categories. This information becomes easier for humans to use and comprehend when it is placed into a visual format.
The general problem is how best to display n dimensions of data on a two dimensional display. The control should allow the user to select the dimensions to be examined. In the real world, the data is often needed in the field or at remote sites. This system is designed to meet these needs over a corporate Intranet or the Internet.
Recommendations for the next iteration
The next step should be to revise the interface based on user input and to add actual database connectivity.
Results of our testing were reasonably positive and supportive of the needs established by company management and our user profile. Therefore, the project should be continued. The next phase should attack three problems:
Additional usability tests should be conducted in thirty days, once the above has been accomplished.