Business Analytics And Big Data Research In Information Systems
For almost a decade, big data and business analytics have been the focus of research and practice. Organizations can generate valuable insight through the combination of business analytics and big data. This interdisciplinary field has a long history of success in scientific inquiry at the European Conference on Information Systems. We offer a synopsis of some of the most prominent themes from the past decade of ECIS’ “Business Analytics and Big Data” track. The synthesis provides a narrative about the field’s evolution and where future research should focus. Three areas are identified that are most likely to be of interest to researchers in the coming years. Each of these areas has its own challenges. This special issue contains six articles. We will conclude by giving an overview and explaining how each article contributes to the understanding of this topic.
1. The European Conference on Information Systems: Past and Present of Business Analytics, Big Data Research, and the Future
Business analytics is the sum of all methods, technologies and applications that are used to analyze past data and to plan and manage business performance. In the past, business intelligence focused mainly on data integration. Because big data can only be used to support these tasks, it is an ideal companion. Big data has brought new challenges and opportunities to business analytics over the last few years. Big data analytics offers enormous potential to provide valuable insight and competitive advantage. However, organisations and individuals are now looking at how to adapt to big data.
As an international forum that brings together information systems experts from all disciplines, the track “Business Analytics and Big Data” has a long and prosperous history at ECIS. The track received 512 applications from 2012, its inaugural year, to 2021. Although the track welcomed research on both business intelligence and knowledge management, it was later split into two separate tracks. ECIS also has other topics-related tracks, such as those related to big data and artificial Intelligence (AI) in certain years. However, this track has been the foundation for IS research in the areas big data, business analytics, big and small data. In recent years, the focus has shifted to data-rich apps and intelligent systems.
The track submissions also show this shift. The track database allowed us to analyse the submissions over the years. We then used text mining to extract keywords from all papers. Marjanovic (Citation2017) and Dinter (2018) have done a similar analysis of 25 years worth of research published at Hawaii International Conference on System Sciences.
The track was initially focused on “knowledge and business intelligence” research. However, the move to the keyword “business analysis” is more than just a change in terminology. Although early topics included data warehousing as well customer relationship management and handcrafted decision-support systems, there was still space for data mining and data science research. The focus changed to more industrial applications of predictive and intelligent analytics and smart objects in 2015. This resulted in a greater emphasis on analysis of large data sets, or big data. These data can be used for everything from industrial maintenance and advertising to process mining and social network analysis. The self-service, cloud, and bigdata management topics emerged. This can also be seen in the diverse topics covered in those years. It was difficult to isolate focal topics (see footnote above).
This shift can partly be attributed, in part, to the increased use of Internet-of-Things, sensors, processing capabilities, and the prevalence of cloud-based, web-based services.
We have seen a shift in how we deal with big data to perform intelligent analyses for various industrial and financial applications. This is a change that takes big data as a given. Many of these analyses rely on machine-learning algorithms that learn from the data and not being programmed. This opens up new research opportunities for future research. The integration of intelligent decision-support systems into human work routines creates new challenges in business analytics research. Many manual decision-making tasks are now automated by novel methods that take advantage of data and automate them.
Latent Dirichlet Alocation of the keywords confirms that these topics have become more of a part of IS research than they were in 2015. This was when IS research began to shift towards analysing large amounts of data and using statistical methods to perform predictive analytics. This same trend is evident in machine learning and AI. Machine learning and AI are now more than just a tool to analyse complex data. These new frontiers offer a range of research questions.
Although some topics may have been altered by changes in co-chairs or additions to the call for papers, or changes in track titles, they reflect the gradual evolution of the community and are therefore still accurate and illustrative of the past decade in business analytics research. It confirms generally the observations of Wanner, Dinter and others. (Citation2022) Marjanovic and Dinter, and continues the work of Wanner et al. (Citation2017 Citation2018). The analysis shows that the field is evolving towards more sophisticated approaches to analysing large and complex data sets. These techniques are now part of the core components of organisations’ competitive business positioning.
2. Perspectives and Opportunities for Business Analytics and Big Data Research in Information Systems Research
These topics are likely to be popular in the future. This list does not include all topics, but it does give some insight into some of the most interesting themes likely to interest IS researchers over time. These topics are a natural extension of the current trends we observe.
Artificial Intelligence for Business Analytics. With ever-growing amounts of data, analytical models that analyse and make decisions rely more heavily on deep learning and machine learning algorithms. Although the models are human-like and allow for decision-making that is human-like, they also have a downside. Humans, as well as data scientists and end-users, may not be able to understand all the models (Janiesch et.al., Citation2021). Additionally, automation of many manual tasks has raised questions about the effects of this phenomenon on workers and society in general. This raises many issues that must be addressed when developing business analytics apps.
– External validity and performance gains: Machine learning algorithms can be used to produce impressive results that surpass established methods in nearly any field. However, only selected metrics must be reported and only one dataset is being used. It is crucial to take into account the generalisability of any intelligent system application in order to make a significant contribution to business analytics research and not just to a specific business practice. The discipline’s progress will be slowed if it focuses only on local optima (Duin and Citation1994; Hutson and Citation2020).
– Governance for intelligent systems: To avoid employee resistance or algorithm aversion, it is essential to carefully plan the operation and maintenance (Jussupow and al., Citation2020).
Transfer learning and retraining strategies are also required to adapt to changing situations (Janiesch, Citation2021). Governance of AI that is based in responsible principles requires many stakeholders at all levels, both within the organisation and across them. It is therefore important to research the future of AI governance (Mikalef, Citation2022).
– Human dignity and intelligent system: How (big) data are generated, stored and analysed for decision-making may lead to both claims and affronts (Leidner & Tona, Citation2021). Business systems can have a negative impact on autonomy, freedom, values, and other personal rights. It is important to continue research on the design of business analytics systems for human dignity.
– Machine learning can reinforce biases rather than reduce them. It is therefore important to understand how humans touch data and the implementation process in order to avoid contamination (Van Giffen, Citation2022). A variety of context factors can influence biases, creating confusion as to which are biased. It is therefore important to investigate how culture affects our understanding of bias and how we can counter it.
– Explanations for decisions: Although every machine learning model can be traced back to its source, intelligent system prediction are often viewed as black boxes. This is because the underlying systems are too complex to understand on your own (Miller, Cited 2019). For the purposes of hybrid intelligence and meaningful decision-making, it is important to understand how explainable AI research can enhance the effectiveness of explanation augmentation. (Dellermann and al. Citation2019). Enholm and colleagues, Citation2021, also discuss how to explain the inner workings (Dellermann et al.).
– Sustainable business models. Responsible AI requires that applications be used in a manner that is not harmful to the environment, society, or individuals. (European Commission Citation2019). Future studies will therefore examine the relationship between AI-based software and sustainable business models. This could be circular economy strategies, for example, or models that place corporate responsibility and social responsibility at their core (Zhao, Citation2018).
Business Analytics: Process mining allows for first-hand insights into how organisations work. With the recent rise and beginning consolidation of process mining tool suites in the market, research into processes mining has evolved from a computer-science-fuelled topic of algorithm and software engineering to a domain-science-induced opportunity to analyse the behaviour of individuals, teams, and organisations based on rich process data.
Hyperautomation: This buzzword refers the rapid, scalable, and measurable automation of as much as possible. Gartner names hyperautomation among the top strategic trends in technology for 2022 (Stoudt Hansen et. al., Citation2021) but it is not yet clear how it can sustainably be orchestrated across operations and measured to business performance (Axmann Harmoko Herm & Janiesch, Citation2021).
– Methods for process mining: Process mining was largely dependent on statistical methods that are part of the data mining realm until recently. Common machine learning algorithms can be used for a variety of purposes. These include image analysis, text analysis, event logs, and process models. This is not the only possible application.
Augmented Business Process Management Systems: These systems combine business process mining, process analysis, and AI technology to produce augmented process-aware processes that can be more autonomous, conversationally-actionable, adaptive, self improving, and explainable. (Dumas, Citation2022). They will be studied in terms of their engineering and contextisation.
– Mining user habits: Process mining relies on event logs (standardised) from process aware enterprise systems. These data are stored by automated tasks that complete user tasks. This data can be used to improve partly digitalized but still manual processes by mining user behaviours (Leno et al., Citation2020).
Governance of Open Data: Big Data not only allows advanced analyses but is also a topic. Data is becoming a more central part of ubiquitous intelligent systems’ decision-making processes. It is therefore crucial to ensure that data can be traceable and confirmed. This includes data required to duplicate findings as well as data used in training analytical baseline models to support further advanced and specialized applications via transfer learning.
– Open Data for Business Analytics: A trend towards open data in science is clear. Many open data institutions, like Dryad, and national initiatives, such as the German NFDI, are fueled by the growing demand for open science data. Even though there are obvious benefits to self-learning systems, open data is not widely available in business. Incentives are also missing. It is becoming increasingly clear that data collected from or generated with public funds should be open to public scrutiny.
– Governance of Data and Analytics: Data storage and wrangling has become more complex in recent times. These activities require new policies, structures, and controls that coordinate and align interests to maximize the potential of data analytics. Analyses must be performed in accordance with regulations. Therefore, privacy-preserving methods to manage data are crucial (Mendes & Vilela Citation2017). Also, access rights as well as data’s role as strategic assets must be thoroughly examined to ensure that organisations can develop data governance practices (Tallon and al., citation2014).
– Entity Linking and Inference of Data: Linking data allows for the creation knowledge graphs to be used in various analytical applications (Abramowicz. Citation2016). Data tying to individuals allows for powerful mass-individualized analyses. These applications offer new storage methods for data, including triplestores and NoSQL database. New opportunities and challenges will be created by the evolving digital infrastructures.
These clusters are reflected in the special issue papers.
3. Special Issue: Research
This special issue of Journal of Business Analytics featured about 12 papers from the track Business Analytics & Big Data at ECIS 2020. They were selected because they had the highest quality papers, best topics, and most interesting discussions. They were extensively updated, enriched, and revised to reflect the changes in conference papers. They were then subject to a rigorous multi-round process of review, which involved both old and new editors and reviewers. Six papers were accepted for the special issue. Two focused on interactions between humans and machine-learning algorithms, two on process analytics and two on data ownership and governance.
Palmer et al. (Citation2022) create a domain-specific mood dictionary and use it to communicate with financial analysts. Their dictionary is superior to other finance-related dictionaries or classifiers, according to their evaluations. Their research has shown that while it may not be superior to a domain-specific and sophisticated machine learning model however, sentiment dictionaries have the advantage and high degree explainability.
Wanner et al. (Citation2022) focuses on the issue of machine-learning algorithms that make it virtually impossible to trace decisions. They use social evaluations to assess the goodness and explainability of machine learning algorithms. They study six different machine learning algorithms. The results of a survey show that trustworthiness is the most important factor.
Andrews et al.
(Citation2022) – Root-cause analysis is performed to identify the root cause of problems with process-data-quality in process mining. The Odigos Root Cause reference framework is introduced by the authors to help in identifying and addressing data quality problems in event logs. It supports prognostic, diagnostic, and label approaches.
Weinzierl et al. (Citation2022) proposes an algorithm that uses deep learning to detect temporal workarounds within business processes. In their open access design science study, the authors devise a method and implementation that can detect seven different information-system-oriented and process-oriented types of workarounds with high accuracy, precision, and recall. Their research bridges the gap between business analytics and organisational routines as well as business process management.
Baijens et al. (Citation2022) describe and theorize governance for data analytics. Their open access paper provides a descriptive framework and a viable-system-model-based view as a theoretical lens to discuss how data governance can create (business) value from data. In three case studies, the authors interviewed 21 companies involved in data analytics. To better understand how data analytics can be managed in practice, the authors created a descriptive governance framework that includes nine sub-routines for structural, process and relational data.
Fadler & Legner (Citation2022) examine the topic data ownership and explain data accountabilities for big and advanced data analysis. They identify three types of ownership based on their four case studies. They present several propositions that differentiate data ownership, product ownership, and platform ownership. Open access is also available.
These six papers demonstrate the breadth of IS research topics. They offer new methodological considerations for governance, ownership and process mining.
