Interactive Visualizations of Big Data

This section is clustered in accordance with the goals of this study. First, the status quo in Austria is going to be presented and discussed (section 4.1) before a deeper analysis of possible barriers is conducted. These barriers are divided into human-related (section 4.2.) and technological-related factors (section 4.3.), as discussed in previous sections.

4.1 Status quo in Austria

4.1.1 Visualization use

For the different visualization types, participants had to answer if the presented types are in use within their companies. "Use" is coded with 1 and "No use" with 0. Answers provided for the various visualization types are presented in Figure 5, which is ordered by the number of visualization types in use. The color code provides additional information and highlights the most common combinations in black and the least common combinations in light gray.

Figure 5. Utilization of visualization types

The most frequently utilized visualizations are business graphics (e.g. line, bar and pie) or type I visualizations, which are applied by 93.8 percent (136 out of 145), followed by geographical visualizations (34.5 percent; 50 out of 145). Common combinations are business graphs with geographical or multi-dimensional visualizations. One noteworthy finding is that 40.7 percent base their analysis solely on type I visualizations. A significant difference between these visualization types can be detected (Kruskal–Wallis test).

The following table presents results on possible influences on the use of type II visualization. We checked whether gender, age, the position held within the company, company size or industries have an effect on utilization. Only gender shows a significant difference. Men use type II visualizations more often compared to women with a difference in means of 0.406.

Table IV Possible influences on use of type II visualizations

	Gender ANOVA	AGE correlation	Position ANOVA	Company size correlation	Industries ANOVA
Type II use	p = 0.040	p = 0.210 r = 0.118	p = 0.427	p = 0.483 r = 0.066	p = 0.103

4.1.2 Interaction use

Analysis on the use of various interaction techniques is presented in Figure 6. This analysis shows that the utilization ranges from 86 answers (67.7 percent) for filtering as the most common technique to 27 (21.3 percent) for the selection of data points as the least common one. Overall, 85.8 percent use at least one interaction technique and most of them use a combination of two interaction techniques.

Figure 6. Utilization of interaction techniques

As type II visualizations are recommended to be used interactively, we check for a correlation between the use of type II visualizations and interactions. Using Pearson's correlation, a significant relationship (p=0.003; r=0.246) is evident; this indicates that practitioners in the field of accounting take note of this recommendation. Additionally, we have ascertained whether interaction use is influenced by any of the variables mentioned in Table V. No significant results can be derived for either of the tested possible influences.

Table V Possible influences on interaction use

	Gender ANOVA	AGE correlation	Position ANOVA	Company size correlation	Industries ANOVA
Interaction use	p = 0.990	p = 0.597 r = 0.050	p = 0.138	p = 0.544 r = -0.057	p = 0.112

4.1.3 Data source use

New technologies in data collection and data storage allow for various forms of data to be analyzed for deeper insights. This section therefore analyses how many and how intensely such sources are currently used. For this analysis, 29 participants did not provide answers and therefore the basis is reduced to 116 participants for evaluation. The analysis presents data on "Use" and "No use," which are coded with 1 and 0, respectively.

Although ERP systems top the list (89.6 percent), they are closely followed by economic data from external databases (81.0 percent). Data that are more likely to be clustered as semi-structured or unstructured such as IoT and Social Media are also quite common. IoT and Social Media are used by 52.6 percent and 53.4 percent, respectively. A notable point is that one third of the participants even use a combination of all mentioned data sources and more than 50 percent use at least a second data source in addition to ERP. In order to test for significant differences between the ranks, the Kruskal–Wallis test has been used for an overall investigation (p=0.000) which is significant.

None of the collected demographic variables show significant influence on data sources in use. Gender, age, and position are not tested because if and how many data sources are used is not influenced by a single person but the whole company (Table VI).

Table VI Possible influences on data sources in use

	Gender	AGE	Position	Company size correlation	Industries ANOVA
Data sources in use	n/a	n/a	n/a	p = 0.354	p = 0.989
Data sources in use excluding ERP	n/a	n/a	n/a	p = 285	p = 0.995

4.1.4 Visualization tools use

With respect to visualization tools, we focus on current top-selling software products. This analysis begins by inspecting the utilization of the various products, which are summarized in Figure 8.

Figure 8. Utilization of software tools used

By far the most commonly used tool in use is Microsoft Excel, which is the basis of analysis for 84.8 percent of the companies represented in this survey (or 96.9 percent if those not providing any answer are excluded). On average 1.5 tools are used, with a combination of Microsoft Excel together with Qlik or Microsoft Power BI constituting the most common examples. Under "other software tools," participants stated, for example, IBM Cognos, SAP BI, MircoStrategy, Jedox, or Infor. The Kruskal–Wallis test indicates that there is a significant difference in usage between the different visualization tool options (p=0.000). With respect to the tested additional variables, only industries show a significant result. A high variety of tools are used in the service industry as well as in wholesale and trade. Industries mainly relying on Microsoft Excel are finance and public administration (Table VII).

Table VII Possible influences on tools in use

	Gender	AGE	Position	Company size correlation	Industries ANOVA
Tools in use	n/a	n/a	n/a	p = 0.885	p = 0.012*
Tools in use without Microsoft Excel	n/a	n/a	n/a	p = 723	p = 0.003**

Notes: *p<0.05; **p<0.01

4.1.5 Summary of status quo

In Austria, more than 50 percent of the participants in the discipline of accounting stated that they use type II visualizations (with geographical visualizations being the most frequently used type) and 85.8 percent indicated the use of interaction techniques to some extent. Filtering, as one of the simplest interaction techniques, is used most frequently. Moreover, the use of type II visualizations is positively correlated with the use of interaction following the recommendation of domain experts to use type II visualizations in an interactive form. Given the fact that we can observe different stages of adaption, we have a solid basis for testing reasons for resistance. These reasons (or barriers) are discussed in detail in the following subsections.

In the context of technical advances, we can observe that the use of various data sources besides traditional ERP systems (representing mainly structured data sets) seems to be common. With an average in data source use of 3.9 and with the inclusion of semi-structured and unstructured data sets such as IoT or social media, it can be concluded that the integration of Big Data into the financial analysis has arrived in practice. With respect to visualization tools, Microsoft Excel is still the most common one, however, other tools are also used quite frequently. Big Data, therefore, is no longer a catchphrase; instead, it has already started to change practices and tools in the management accounting profession. Interestingly, the use of tools besides Microsoft Excel is significantly influenced by industries. While companies in the service industry (including advisory) have a high rate of adoption, companies in traditional finance domains (banking, insurance) seem to resist the use of other visualization software tools such as Microsoft Power BI or QlikView.

4.2 Human-related barriers

4.2.1 Resistance to new visualization types

To answer the first hypothesis, we analyzed the difference in use between interactive type I and interactive type II visualizations. Consequently, we separated our data file by those at least using one kind of interaction and those not using any interaction at all. In sum, 109 participants indicated to use interaction, which is the basis for our comparison of interactive visualization type use. Thereof, 41 participants indicated that they only use type I, two participants stated that they only use type II, 62 use a combination of type I and type II, and four use none of the presented visualization options (most likely they are using tables with filtering options). Significance testing based on the Mann–Whitney U test shows that interactive type I visualizations are used more often than interactive type II visualizations (p=0.000), rejecting our null-hypothesis stating no difference between their usage. Therefore, we can detect a resistance to change when it comes to the adaption of newer and probably unfamiliar types of visualizations.

4.2.2 Resistance to new interaction techniques

The logical next step is to analyze if users are also persistent to interaction techniques. Again, the basis for calculation are the 109 participants stating to use interaction. Thereof, only three indicate to use solely advanced interaction techniques, 61 indicate to use solely simple interaction techniques and 45 indicate to use a mix of both. The difference based on Mann–Whitney U-test shows that simple interaction techniques are used significantly more often than advances ones (p=0.000). Based on these results, our null-hypothesis (H3) can be rejected.

H3. There is no difference in use between simple and advanced interaction techniques.

4.2.3 Perceived EoU

To test H2 and H4, perceived EoU is measured. The hypotheses propose a correlation between the use of multiple type II visualizations as well as the use of multiple interaction techniques and the construct's perceived EoU. The correlations are calculated using Pearson's correlation coefficient, while Cronbach's α was utilized to test the internal reliability of the construct. Cronbach's α of perceived EoU is 0.767 and therefore above the 0.7 threshold. The mean level of agreement of the four questions described in Section 3.1 lies between 5.53 and 4.52, which is well above average. This indicates a medium to high perceived EoU for interactive visualizations. For correlation analysis, a sum score of the four constructs is used with the results presented in the following table (Table VIII).

H2. The lower the use of type II visualization, the lower their perceived ease of use (EoU).

H4. The lower the use of interaction techniques, the lower their perceived EoU.

Table VIII Results of perceived EoU

	Type II count	Interaction count
Perceived EoU
Pearson	0.279**	0.220*
Sign.	0.002	0.018
n	116	116

Notes: *p<0.05; **p<0.01

Results indicate that the higher the use of type II visualizations, the more likely participants are to perceive them as helpful in their daily working experience. Furthermore, the strength of the impact of usage on perceived EoU, with a factor of 0.279, can be classified as moderately strong. The same seems to be true for the use of multiple forms of interaction and therefore H2 and H3 can be confirmed.

4.2.4 Familiarity

Visualization types were rated according to the participants' familiarity on a seven-point Likert scale. We included familiarity with visualizations in our study as visualization types could still be known even though they are not used. 1 represents no while 7 indicates a high familiarity. Results based on ANOVA and a post hoc SNK are presented in the following table (Table IX).

Table IX ANOVA familiarity with different chart types (seven-point likert)

ANOVA and post hoc SNK	1	2	3	4	5	6	7	Average	Significant sub-groups		Industries ANOVA
Business graphics	0	3	1	1	2	22	116	6.669				1.000
Geographical Vis	6	10	15	16	43	36	19	4.821			1.000
Multi-dimensional Vis	19	29	35	18	27	14	3	3.986		1.000
Text-/Webbased Vis	14	19	23	21	43	17	8	3.407	0.210
Network Vis	25	26	36	25	23	8	2	3.186

Notes: *p<0.05; **p<0.01

This analysis demonstrates that type I visualizations are by far the most familiar visualization types, which is in line with the high utilization presented in the previous analysis. Based on these results we can reject H6 indicating no difference in familiarity between type I and type II visualizations. However, based on the aggregated average of all type II visualizations it seems that the majority of the participants have at least some experience with type II visualizations. Only five participants indicated that they are not familiar at all (the average score for type II visualizations is 1) and 10 indicated that they are mostly unfamiliar with them (average score of type II visualizations is 2). 49 participants indicate a familiarity above average (above 5).

Again, we test for possible influences of the variables collected (gender, age, the position held within the company, company size or industry). Significant results can be obtained between industries. In service and public administration, we identify a high familiarity, while for the transportation, communication and electric industries a low familiarity is evident. In addition, there is an indication of a higher familiarity depending on positions. Participants in higher positions (top or middle management) are more familiar with type II visualizations compared to participants in lower positions (lower management or employees) in management accounting. The results are presented in Table X.

Table X Possible influences on use familiarity with type II visualizations

	Gender ANOVA	Age correlation	Position ANOVA	Company size correlation	Industries ANOVA
Familiarity_Average	p = 0.406	r = -0.034 p = 0.722	p = 0.077	r = -0161 p = 0.088	p = 0.045
Familiarity_Sum	p = 0.398	r = -0.047 p = 0.621	p = 0.074	r = -0.154 p = 0.103	p = 0.076

To test H5a and H5b, an analysis of the correlation between familiarity and type II utilization as well as familiarity and the perceived EoU has been conducted. The results show a strong positive correlation for both usage and perceived EoU. Therefore, both hypotheses can be confirmed.

H5a. The lower the familiarity with type II visualizations, the lower their use.

H5b. The lower the familiarity with type II visualizations, the lower their perceived EoU.

Table XI Results of familiarity

	Type II count	Perceived EoU
Familiarity_Average
Pearson	0.340**	0.350*
Sign.	0.000	0.000
n	145	116
Familiarity_Sum
Pearson	0.330**	0.372**
Sign.	0.000	0.000
n	145	116

Notes: *p<0.05; **p<0.01

4.2.5 Summary of human-related barriers

Based on this analysis, we can state that a medium degree of familiarity regarding type II visualization is already present in practice. However, only if type II visualizations are used as intended (in combination with interaction techniques) can they release their full potential and enable users to benefit from their use. The lack of willingness to deal with more advanced interaction techniques negatively affects the use of more complex type II visualizations. It is necessary to increase the familiarity for both type II visualizations and advanced interaction techniques in order to achieve more widespread usage throughout industry sectors. As soon as this initial barrier is crossed and participants are familiar with type II visualizations, the perceived EoU will also be positively influenced and thus frequency of use will be enhanced. This last part is essential as it indicates that type II visualizations are not dispensable, as they are considered useful by those knowledgeable. The barrier lies in introducing new options to their user base in an appropriate manner.

4.3 Technological-related barriers

4.3.1 Data sources

Semi-structured or unstructured data sets are mainly connected to economic data, web analytics, social media data, and sensor data, while structured data sets are related to traditional ERP systems. In this analysis, we want to check whether a high usage of semi-structured or unstructured data sets correlates with the likelihood of turning to type II visualizations as well as to a higher number of interaction techniques.

The results in Table XII indicate that a moderate correlation for type II visualizations can be found, while for interaction no correlation exists. Therefore, H7 can be confirmed while H8 needs to be rejected.

Table XII Results on data sources

	Type II count	Interaction count
Source Count
Pearson without ERP	0.236**	0.179
Sign.	0.011	0.055
n	116	116

Notes: *p<0.05; **p<0.01

H7. The lower the number of various data sources, the lower the use of type II visualizations.

H8. The lower the number of various data sources, the lower the use of interaction techniques.

4.3.2 Visualization tools

Analyzing for a relation between the number of visualization tools and the number of different interaction techniques and type II visualizations is done using Pearson's correlation. While there is an effect for the number of tools in use with respect to type II visualizations, there is no effect with respect to interaction count. Therefore, H9 can be confirmed, while H10 needs to be rejected. This analysis is additionally calculated with and without the integration of Microsoft Excel, as Excel does not provide sufficient support for either type II or for advanced interaction techniques.

H9. The lower the number of visualization tools used, the lower the use of type II visualizations.

H10. The lower the number of visualization tools used, the lower the use of interaction techniques.

Table XIII Results of visualization tools

	Tools count	Tools without Microsoft Excel
Type II count
Pearson	0.330**	0.345*
Sign.	0.000	0.000
n	127	127
Interaction Count
Pearson	0.132	0.141
Sign.	0.139	0.113
n	127	127

Notes: *p<0.05; **p<0.01

To further conduct our analysis based on tools, we also explore those solely basing their analysis on Microsoft Excel. Using ANOVA (the split variable is sole Microsoft Excel compared to a combination of tools in use) to calculate the difference in the number of used type II visualizations, it is evident that there is significantly less usage for those participants using only Microsoft Excel (p=0.000). The sole use of Microsoft Excel can therefore itself be seen as a barrier. No results can be obtained based on interaction techniques.

4.3.3 Summary of technological-related barriers

In the context of technological-related barriers, we can first and foremost identify the sole focus on Microsoft Excel as a barrier. The use of different data sets, which are semi-structured or unstructured in nature, can be identified as an enabler or driver. The more data sets that are integrated into traditional reporting and management information systems, the higher the likelihood of type II visualizations being employed. The same can be said about visualization tools as their use also increases usage of type II visualizations.

With respect to interaction, no correlations can be found between data sets or visualization tools. More advanced interaction techniques are not even applied if data sets require or visualization tools offer their use.