Impact of data visualization on decision quality PROBLEM DEFINITION 1. 1 Purpose of the study The purpose of the study is to research the effect of data visualization on decision quality. 1. 2 Context of the study Visuals have been used for centuries, saying a picture is worth more than a thousand words proves this. From geographical maps to stock prices, data visualization is widely used to represent information visually. In BI, due to the increasing trend into data collection and other data mining techniques, 3D charts and dashboards among other visuals (Duke et al. 005; Lester n. d. & Blewett n. d. ), have been widely accepted as important tools in business analysis and have aided in companies maintaining some competitive advantage by making quick and accurate decisions. Data visualization therefore can be seen as an important part in the decision making process. How then can businesses make sure that the decisions they make are of a high quality? This research will focus on the data visuals and fill the gap in research as well as practice. 1. 2. 1 Data mining
Data mining uses mathematical formula in constructing models(Davenport et al. 2001), which would lead into establishing patterns(Viktor 2009), in mapping out business decisions and strategies. This has lead to the extensive widespread use of data mining in marketing, medicine, finance. It is the production of these patterns that make data mining techniques very crucial, though huge data sets cannot always produce results difficult to understand. Watson (2009) highlights the numerous benefits of BI, as well as pointing out how difficult it is to measure them.
Though providing examples, of only successful implementation, it fails to reflect on possible weaknesses of BI thereby missing on lessons to learn from failed implementations. In conclusion, as BI becomes more pervasive so does the cost of training and support. 1. 2. 2 Decisions and decision-making process Decision theories are very conflicting and could be classified into normative and descriptive (Clark 2010) and (Keren & Bruin 2003), and judged by the process or by outcome(Keren & Bruin 2003). Decision quality is always a hindsight action, and as such makes judging it even more difficult and complex.
Clark (2010) though full of practical examples contains no empirical data to support its claims. This paper is written in a marketing manner than an academic paper. Keren & Bruin (2003) though well written and containing numerous practical examples and simulations, has a lot of statistics. This paper however, could be used in selecting a weighing method to use. These two papers do provide more insight into the complexities of decision quality and try to shed light, into which methods are more accurate than others.
Focus though tends to be on the decision analysis and less on decision quality Due to myriad definitions and the fact that it is a complex subject, which could be broken down into a systematic and concise manner (ibid. ) Judging the decision quality could be viewed from two angles, one being the decision maker, secondly the judge. Both present their own problems, the decision maker has to deal with anchor traps, status quo traps, sunk cost traps and confirming-evidence traps among others (Hammond et al. 1998).
The judge on the other hand might choose a different model to the decision makers’ thereby wrongfully assessing the decision quality(Keren & Bruin 2003). The judge assesses using three methods: gambling paradigm, conflict method and accountability method (ibid). Numerous statistical weighing methods could be employed (Jia et al. 1997), though practice and theory does yield differences. In general, people do not really care how decisions are made but as to the quality of these decisions, and moreover people focus mostly on outcomes ather than the process itself (Keren & Bruin 2003). An example provided is that of a doctor conducting surgery on a patient. How is this assessed? However, understanding how decisions are made could help influence better decision-making. Davenport et al. (2001) found out that most decision makers base their decisions on unrelated factors to the data analysed and or collected. An alternative option, consider both the process and outcome in assessing decision quality. Decision and result are two different things, something normally ignored.
Findings and decisions themselves do not yield results. A decision needs to be implemented in order to yield results (ibid. ). Time elapses between when a decision is made and results happen (Keren & Bruin 2003). Good decisions can therefore lead to bad results. Some companies base their decision quality on financial results (better profitability, revenue etc. ). Cost saving seems to be the best indicator to measure and predict as far as financial results are concerned. However, result is irrelevant as a measure of decision quality when latency issues are taken into consideration.
These papers though proposing methods and ways of determining decision quality, do little in as far as providing quality measurements. When would it be good? (Watson 2009) though highlighting the importance of Decision Support Systems, like most other papers, disregarding disadvantages such as obscuring responsibilities. This is a situation whereby people deflect responsibility and blame it on the computer making the decisions. 1. 2Definitions Lack of consensus, prevalent in IS/IT, on definition of terms and concepts seems to hamper progress.
There is the unresolved issue of what constitutes data, information and knowledge (Tuomi, 1999). Definitions of data visualization seem to vary depending on its context. Search Business Analytics (n. d) define it as “general term used to describe any technology that lets corporate executives and other end users “see” data in order to help them better understand the information and put it in a business context. ” However, the word “see” could be viewed as being rather ambiguous. In Business Intelligence, data visualisation is the use of visual epresentations to explore, make sense and communicate data (Viktor & Paquet, 2009). This raises the question of whether all data presented visually is only a form of data visualization? The key to answering this question would be to look at data analysis, which deals with discovering and making sense of data. The goal of data analysis therefore is to make sense of the data and as such assist in achieving a decision based on that understanding. A pragmatic viewpoint could be to assume; if data or information were represented in a format that aids decision-making then the term data virtualization would be adopted.
This is because it is possible to have graphical representations, which are not used for decision making but as such informative. Data Visualisation issues Data visualization makes it easy for this information to be made, colour instantly shows similarities and differences. The role colour plays seems to be an overlooked property in data visualisation. Many modern data visuals are flawed in their design and convey the wrong information and lead to incorrect information being passed on from the designer to the user and therefore not being effective in the decision making process(Eckerson & M.
Hammond 2011). This does have massive social, political and financial consequences to businesses and people and sometimes not recognised. Gershon et al. (1996) point out that images too have disadvantages and at times words are more effective, while Bresciani & Eppler (2008) highlight misleading perceptions of their reliability, implicit meanings inherent, and high prerequisites for diagram interpretation. Source of these errors could come from the designer or from the user, and could be intentional as well as unintentional.
Businesses should as such focus less on analytical tools but on methods that deliver maximum benefits. With this in mind, an excel table might be adequate for an executive but not for a business analyst where lower levels of abstraction of data is needed. A balance of imagery and text can sometimes be all what is required. Data visualization is not “plug and play solutions” (Eckerson & M. Hammond 2011); something which businesses need to be aware of . Instead the role of the user, purpose, and organizational culture are much more important (ibid. ).
Though these papers lack any empirical data to back up these claims, they are influential in laying focus more on the people using these techniques and less on the software. They can therefore provide ground to further research into this area. 1. 3 Problem Statement Focus on the tool by companies’ means that it is difficult for them to actually assess whether or not the data visuals they base their decisions on are improving their decision quality. There is a clear gap in the literature on data visualisation with regards to user experience and usability, as well as how this affects the quality of decision.
Current available literature in this field will be discussed in the next section. 1. 4 Significance of the study The study fills a gap in that it is an empirical study into data visualization, within business intelligence and decision quality. Harris & Davenport (2005) and Blewett (n. d. ) observed that businesses are drowning in huge amounts of data, affecting speed and efficiency of decision making, as most decision-making is still a human task. ROI of implementing and investing in BI are therefore not fully realised, as companies are rather overwhelmed.
Since companies are not focusing on the end product (usability) and the end user (user experience); designers and end users lack necessary methodologies and techniques to enable their decision making process(Chen 2005). After all, it is people that use these reports to formulate strategies. Chen (2005) though highlighting issues does not go further to establish any possible working methodologies nor does it provide any empirical evidence, but criticises others. The paper however builds a case for more usability studies in visualisation tools.
Development in software could make all human involvement obsolete through automated decision-making as discussed by Harris & Davenport( 2005); a paper sponsered by Accenture could be viewed as being baised. Interestingly, automated systems though cannot handle actions or decisions that are exceptions to the rule. Automation also raises staffing as well as disregard for the end users in the organisations. The research will be of importance to any stakeholders within Business Intelligence. Practioners invloved in making and using data visualisations in formulating decisions could benefit.
Companies could improve their processes by noticing omitions in their own processes. New comers to Business Intelligence could use this research to guide them into formulating processes that would improve their decision quality. This research could also benefit the research community as it could increase research into other factors that influence decision quality and more focus on user experience and uability issues in Business Intelligence, with a user centric focus. 1. 5Deliminations of the study
This research will focus on using existing visualisations tools and applying known design techniques. 1. 6 Aim The primary aim of this research is to show the effect, if any, data visualisation has on decision quality. 1. 7Research question How can data visualization affect decision quality? 1. 8Objectives -A critical analysis of current literature pertaining to data visualization; with emphasis on business intelligence. -Human perception and human cognitive behavior with regards to graphical representation; the Gestalt psychology and how that could assist in business intelligence Measuring decision quality; select an appropriate method that will be adopted in this research. -Formalise a classification system that be adopted in categorising graphical representation. -Use of colour and role colour plays in data visualisation -Compose some guidelines using methods and techniques investigated which assist in shaping good quality decisions. -Carry out experiments using guidelines, with a view to assessing the role data visualization plays in decision quality. -Collect data from above mentioned experiments and evaluate data collected. References: