Interactive Visualizations of Big Data

The data collection was completed by the end of 2016 with an online survey. The questionnaire was sent out as well as distributed to participants at the “Controlling Insights Steyr” event, where 337 business practitioners and leaders from 192 different companies took part. This event annually gathers managerial accountants within Austria and therefore provides a solid basis to analyze the status quo of the reporting practice within this area. From the 337 attendees, 105 participated in the survey. To increase the data sample, the questionnaire was additionally sent out to alumni of an Austrian economic university, namely, Facebook. The university has a study program specifically designed for managerial accounting and hence includes the target audience needed for this analysis. This two-step sampling approach resulted in 145 evaluable responses from a broad variety of business sectors.

3.1 Questionnaire

The questionnaire started by introducing the purpose and also by depicting various visualization types under investigation. The visualizations used are clustered by type I (business graphics encountered in everyday life) and type II visualizations (geographical, hierarchical, multi-dimensional, network, text and geographical visualizations). With respect to these options, participants had to answer whether those visualizations are in use. Additionally, they rated familiarity with each type on a seven-point Likert-scale.

For collecting information on interaction techniques, participants were asked to indicate their use. Classifications are made based on the discussion in the theoretical background and other widely cited studies. The following interaction techniques, which are clustered by simple and advanced techniques, have been used:

1. Simple interaction techniques:

   • assigning data to axis (drag and drop, drop-down);

   • filtering the data; and

   • assigning color and symbols to data (brushing and highlighting).

2. Advanced interaction techniques:

   • selecting data points for further analysis (zooming, drill-through); and

   • multiple views of the same data.

For an indication, if interactive type II visualizations are a useful concept, we used the construct of perceived EoU from the technology acceptance model introduced by Fred D. Davis in 1989, which is regularly cited and well recognized in the information systems literature. Per this definition, perceived EoU is “the degree to which the prospective user expects the target system to be free of effort”. More precisely, the participants had to indicate on a seven-point Likert scale if:

• interactive visualizations are easy to understand;

• they support the comprehension of content;

• they decrease task difficulty; and

• they increase working performance.

By analyzing these questions, we have been able to derive indications on the perceived benefits of interactive type II visualizations. In the following, these can then be contrasted with the costs of adaption.

With respect to the tools used, options based on frequently used tools of the Gartner Magic Quadrant 2016 were chosen: Qlik, Microsoft Power BI, Tableau, R, SAS and in-house developed software. In addition, an option for others was presented to the participants. For the list of data sources offered, we referred to the IBM report. The options given were ERP, economic data, geographical data, web analytics, social media, and sensor data (IoT). Additionally, we provided the option others.

Finally, information on demographic information was collected. Demographic data results are summarized and presented in the next subsection.

3.2 Demographic information

Table I provides an overview of the respondents, clustered by business sector, Table II summarizes the positions held within the company and Table III  summarizes age as well as gender. The tables demonstrate numerous business sectors and therefore support the generalizability of our results across industries. The high proportion (about 50 percent) of participants in management positions indicates high quality data.

Table I Overview of respondents, clustered by business sector

Table II Overview of respondents, clustered by positions

3.3 Data analysis

For data analysis, we coded answers with Microsoft Excel and carried out the statistical analysis using SPSS. To evaluate visualization and interaction use as well as data source and visualization tool utilization, a graphically modified table provides information about which and how many are currently in use. Subsequent to the descriptive statistics, the Kruskal–Wallis test was applied in order to perform significance testing. Furthermore, a variance analysis (ANOVA) and Pearson correlation was conducted to determine differences between groups or correlations respectively. Details on the conducted analysis are also provided in the respective results section.

3.4 Limitations

Due to the data collection with a standardized survey, there was no possibility of dealing with participants individually, which could lead to a misinterpretation of the given answers. To proactively avoid a misinterpretation of questions, we attempted to phrase all questions unambiguously and provide introductory information about the purpose of the survey. Pre-tests and interviews with five participants were conducted before launching the study. Further, the sample population is Austrian, which might cause issues in terms of generalizing results to other geographical areas or cultural heritages.