A Survey of Visualizing Business Data

This section exemplifies literature that places emphasis on a complete business ecosystem from an external perspective without the specific focus on a single body within the ecosystem. These ecosystems can be collaborative, competitive environments, industries, or stock market-based. The focus is often placed on an overview of the complete ecosystem. This classification is dominated by research from Rahul C. Basole, who specialises in Business Ecosystem visualisation, and is associated with eight papers of the seventeen in this field.

Primary Data as Intentional, Active Research Study (BE): Here, we present a visualisation paper that performs a research study using business ecosystem data. The ecosystem is represented by stock market activity. Merino et al. presented a user study evaluation of different visualisations and identify which are more suited to visualising large amounts of stock market data. This is done through the "Task-At-Hand" interface which offers a selection of visualisation techniques to display the data and user options that incorporate brushing and linking.

Traditional visual designs range from bar charts and line graphs to pie charts and tables. These visualisation techniques are easy to understand but are limited in the depth of data that they can display. The Geometrically-transformed group contains visualisations such as parallel coordinates and pyramid representations of data. These are useful for identifying trends in the data but, again, it can be challenging to interpret more complicated datasets.

Iconic displays map attributes for complex datasets. Glyph-based approaches depict a large number of data attributes, however, they are not appropriate for large datasets due to overplotting. Pixel-based designs depict the benefit of conveying a large number of data points but present a challenge when positioning the data points being plotted. Stacked Display techniques are designed to present hierarchical data. Treemaps are often used for visualising data of this nature. The space-filling approach enables useful analysis of a dataset; however, it is dependant on the algorithm used to generate the treemap as varying aspect ratios can have a negative impact on utility.

Each design category is assessed by a range of criteria. See Figure 13 for results. The visualisation technique categorisation was identified by Keim and the criteria by which the designs are judged were identified by Zhou et al.

Figure 13. The results table for the task evaluation. The "Charts" category scores highly, whereas the "Parallel Coordinates" scores relatively poorly. Evaluation criteria such as "Correlations" only scores highly in the "Charts" category. This shows the limitations of more elaborate visualisations, and the value of simplicity. Image courtesy of Merino et al.

Basole et al. performed a study that evaluates the effectiveness of visualisation methods on business ecosystem data. The three presentation types examined – list, matrix, and network – are the most popular forms of ecosystem visualisation. Seven criteria were used to evaluate the visual representation: ease to learn, ease of use (beginner), ease of use (intermediate), speed of use, speed to learn, control over analysis, and flexible capabilities. See Figure 14.

Figure 14. The results of the study. Network outperforms the other two methods, but the matrix method provides a consistently quick and useful visual tool as well. Image courtesy of Basole et al.

The results found that the network outperformed the other two in almost every criterion, other than the ease to learn. Once the cost of training has been expended, the network is a far superior method of displaying business ecosystem data.

Hybrid Web-scrape (BE): In the following category, web-scraped data are used to visualise the business ecosystem. Lu et al. outlined methods that use visual analysis to predict box office success. Social media data are used to derive predictions, ranging from Twitter data to Bitly link data mining.

To obtain customer data about the movies, tweet data are mined using keywords related to a given movie. In addition to this, IMDb is used to collect numerical data about the film. All data are collected two weeks before the release date. Both the volume of tweets and the content are taken into account when analysing the tweet data. Sentiment analysis is performed on the text so that the film reception can be assessed and the qualitative data are quantified. A word cloud of the most commonly used words where colour is mapped to sentiment enables users to see at a glance how positively the twitter community ranks the film. A timeline visualisation shows the magnitude of positive or negative tweets over a two-week period before the release of the film. This enables marketing campaigns to be measured through a new medium. Both film review scores and the film revenues were predicted from the data. Despicable Me 2 has a predicted score of 7.8 and an actual score of 7.9, whereas the predicted five-day revenue for the film is $116.5 m and the actual five-day revenue is $143 m. This shows the score is far easier to predict than the revenue, but the system can calculate a rough estimate.

Prediction models for the film industry have previously been worked on. The relationship between movie review and revenue is also examined. Asur and Huberman found that the volume of tweets relating to a film had a direct relationship on the revenue, accounting for 80% of the variance in prediction.

Secondary Data as A Priori Database (BE): The literature in this section contains a number of visualisation research papers on the business ecosystem where the data are taken from pre-existing (a priori) databases. This is by far the most common data source for an ecosystem visualisation paper. Wu and Phillips presented a visual design of the 2008 financial crisis. Using stock prices and news headlines on a timeline leading up until the crash, the user can identify relationships between financial news headlines and the Dow Jones Industrial Average (DJIA).

The user is presented with a dashboard-style visual design that uses brushing and linking techniques to show different aspects of the finance evolution. The dashboard is made up of three components. The bubble motion chart depicts the influence of news articles on the trading of stocks. This visual design projects a live animation to present the data. A radial plot displays the frequency by which important words are used in the headlines of financial news articles. Words that result in a positive impact are drawn in green. The News Events Bar Chart shows a simple chart that indicates the number of news articles written about each of the chosen companies up until a given point.

Lux presented the first overview of financial visualisation, which Wu and Phillips extended. Merino et al. analysed different visualisation techniques for financial information to discover those most suited to the data. Keim et al. suggested that charts are the most efficient method of presenting financial data.

Basole et al. presented an overview of the business ecosystem using the example of the mobile industry. To achieve the overview, Basole et al. presented dotlink360, an in-house visual analytics application that utilises data from pre-existing finance and business databases as well as current news articles to explore a businesses' ecosystem. This involves exploring the connections between companies and the types of connections, the difference between companies in similar market position, and how these positions have changed over the years. See Figure 15.

Figure 15. The dotlink360 application. This network clusters firms by industry. The position of the nodes is based on the weight of their involvement within each of the industries listed around the perimeter. The user can select a firm to see their involvement with other companies within their ecosystem. Image courtesy of Basole et al.

Previous software that attempts to yield similar insight into the business ecosystem does not take into account the dynamic and complex data involved. This research takes a more holistic approach to the data analysis which encompasses the complex data landscape "from core to periphery". Emphasis is placed on the user interaction design to ensure a practical use in the field. The dotlink360 incorporates a range of visual designs, namely a "periscope view" that maps the companies in a network that are clustered by industry.

Few citations to previous related work are provided here, most notably Basole's visualisation of interfirm relationships in a converging ecosystem.

In a later paper, Basole et al. also presented methods of visualising relationships between firms within an industry that enable analysis of the surrounding business ecosystem. The networks that link businesses together are the focal point. By extension, the designs convey a holistic competitive insight into the ecosystem beyond the traditional single market view. See Figure 16.

Figure 16. The Path view in the connectivity perspective of the software. Clusters of companies can be seen as market segments where the links between nodes signify the company agreements. Image courtesy of Basole et al.

The software primarily focuses on the agreement portfolios of an entity, defined as either a company, market segment, or country. An agreement is classed as an official interaction between two entities whereby a decision has been processed.

The connectivity perspective is the primary window that shows the set of connections between firms. See Figure 16. There are four types of view in the connectivity perspective: Path, a network view that shows connections between companies; Segment, a view that depicts a company's position relative to its market segment (company focused); ScatterNet, a node-link diagram combined with a scatterplot to show company-to-company agreements; and Geography, a view that maps the physical location of the companies on a map.

The foundation of research in this paper stems from the development of network analysis software that shows static imagery of organisational networks. The business process analysis component of the research has a foundation in previous business visualisation tools. Knowledge management and discovery tools were also available but in separate software packages.

Schotter et al. explores international business relationships through the creation of interactive dashboard designs. Data are overlaid on a google map view which represents the pathways in which Japan is expanding their business throughout Asia. This interactive map shows the "communities" within a network, where nodes share a higher rate of connections than in other areas. Additionally, heat maps and hexbinning (tile-based heat maps) are also used to show the relationships between these communities.

The paper presents a range of visual designs that utilise the plotting of meta-data over a google map window, or through standard visualisation methods such as matrix plot and chord diagrams. These dashboard-like designs focus on the nature of the relationships between business entities, and the communities within an ecosystem.

Basole et al. presented ecoxight, an interactive network diagram displaying business ecosystem data. The network is constituted of two components, nodes and edges. Nodes represent either a company, API, or investor within an ecosystem and edges which represent the relationships between the nodes. Each edge has a source, target and weight. User options enable interaction through data filtering and visual control of size, colour, and shape of nodes within the network. Multiple views enable the user to explore the data from different perspectives.

Secondary Data as a Business Process (BE): This subsection features two visualisation papers that focus on the business processes involved in the business ecosystem. Both use data that were not collected originally by the visualisation researchers (secondary data). Basole utilised visualisation tools to enhance understanding of a business's position in the global market by exploring supply chain network process. The approach in creating the tool is to co-author the design with the corporate users to ensure the use case is being adhered to.

Three scenarios are devised and tools are created to cater to each of their user requirements.

The global supply chain is an integral process to the success of a business. Poor optimisation of this process can wreak havoc on the finances of the business in the future. The tool depicts the risk involved in managing a global supply chain using a network graph to show suppliers and the relationships between them. Shared suppliers are seen as a higher risk and are therefore highlighted. Through this visual design, we can clearly see what suppliers should be used to minimise the risk of chain disruption. See Figure 17.

Figure 17. The force-directed network visualisation of ecosystem convergence. The nodes represent market segments and edges show inter-firm relationships between the segments. Image courtesy of Basole.

It is essential for corporate management to have a detailed and accurate view of the competitive market in which they are engaged. A force-directed network is used to achieve this. A user can view the market ecosystem as a whole which enables informed decision making. Venture capitalists are highly sought after in the business world. Understanding the motivation behind the investments these people make is integral to gaining funding from them. Here, the previous force-directed network is adapted to show venture capitalist activity within those markets.

Previously, corporate visualisation has been studied, but no focus is placed on intelligence tools.

As business globalisation increases, the process of supply chain management is becoming increasingly more complex. Large supply operations are often subject to delays and susceptible to extraneous environmental factors outside business control. Basole and Bellamy provided insight into visual forms of risk analysis for supply chains. Network graphs are used to depict the risk across all supply chain visualisations. There are two main focus areas: firm level and industry level.

For firm-level analysis the network graph places the firm at the centre of the layout and all firm connections are shown within a circle radius connected by relationship curves. See Figure 18. The connecting firm's dependants are then shown on the next circle radius. This continues until a map of the supply chain is built around the first firm. The industry level visualisation places all firms in an industry as nodes around in a circle. Relationship edges connect the firms and colours are used to map the risk level. Node size is linked to topological importance.

Figure 18. A three tier supply network of a major electronics manufacturer. The inner layer represents the primary supply dependants of the manufacturer, the middle layer represents the secondary dependants (primary dependants for the inner circle of dependants), and the third layer represents the tertiary dependants (primary dependants of the middle layer of dependants). Image courtesy of Basole and Bellamy.

Decision support related research has been carried out in the field of network topologies. There is previous research investigating workflow visualisation, identifying influential nodes in topological networks, and identifying unhealthy supply networks.

Secondary Data as a Business By-Product (BE): In the following, business by-product data are used to depict the business ecosystem. Business by-product data often come in the form of transaction and sales records. Ko et al. created a visual analytics system that draws a comparison of two competing businesses, displaying trends, growth rates, sales, etc. The software utilises PoS (Point of Sale) data to create the analysis and design. Visual features of MarketAnalyser can be seen in Figure 19.

Figure 19. The main window of MarketAnalyser. The left panel contains all the filtering options and colour legends. The main screen space conveys the pixel based comparisons, and the right ride panel contains the geographical view. The bottom panel contains the stacked bar view and time sliders. Image courtesy of Ko et al.

A screen space saving layout is implemented to display sales, growth, and trends for each of the chosen companies. Filtering options enable the user to prioritise stores through a cumulative method of filtering companies using sales or trends. The user can choose the most suitable company to compare with their data. Figure 19 on page 32 shows the geographical view of the software. This enables the user to see the geographical sale locations for each company. The colour of each region shows the direction of the trend in sales. The stacked graphs in Figure 19 can show relationships between different product purchases. It can highlight combinational trends of multiple products between businesses. Sliders are used to select time intervals.

Market trends and financial forecasting are traditionally conveyed using the standard set of visualisation tools such as bar charts and line graphs. Treemaps are sometimes used to represent market data. The design of the display matrix utilises elements from Keim's work on pixel-oriented visual designs.