Business Intelligence

Rapid, cutting-edge developments in intelligence technologies and methodologies over the last decade give the impression that business intelligence (BI) is novel. However, business intelligence systems have been transforming organizations for decades (Nersessian & Mancha, 2021). Reduced cost of storage coinciding with massive interest in technology enabled decision making are a few of the drivers of recent phenomenal innovation and growth in BI technology. Organizations are extremely motivated to invest in intelligent systems as globalization has resulted in compressed business cycles (Sharda, 2020, p. 79). Insight from data can drastically enhance speed and proficiency of managerial functions, potentially unlocking impressive competitive market advantages. As a result, organizations are increasingly looking to data to drive decision making, reducing uncertainty associated with traditional management techniques.

Business Intelligence Strategies

All business intelligence ventures aim to provide leaders with the right information, at the right time, in the right place to optimize decision making. Organizations collect, store, and maintain sets of structured and unstructured data generated by a variety of technology supported interactions. These data can potentially contain a myriad of details about transactions, costs, market history, human demographics, churn, and media exchanges, to name a few. Yet, these data often provide little value beyond historical documentation without processing. Analysis and subsequent business insight is facilitated by organizing and transforming stored data.

Business intelligence objectives must align with business strategies. Organizational strategic plans outline specific long-term objective statements defining how mission and vision is achieved. Evaluating objectives to determine business intelligence applications requires analysis of what needs to be done, what is being done, and what could be done. Measures are developed to quantify proficiency for each action category. Objectives are then examined by decomposing each statement to determine operational purpose, followed by identification of business functions contributions to accomplishing the goal. Comparison between what needs to be done and what is being done within each function reveals potential opportunities for future application of business intelligence. Modeling phases delve deeper into issues to produce problem statements, which serve as the basis for developing and validating multiple business intelligence models.

Forming SMART goals can help outline a path for business problems. It is important to note that problems should be evaluated extensively to determine if business intelligence is the appropriate solution.

Business Intelligence Processes, Tools, and Technologies

The foundation of business intelligence is understanding the important elements involved in making decisions. This can be boiled down to four steps: define a problem, construct a model, identify, evaluate, and compare solutions, then determine the best solution. Analytics are used to develop appropriate models in to address business problems and opportunities. Business analytics are generally described as either descriptive, predictive, or prescriptive. Descriptive statistics examines historical data to determine what has or is happening. The output of this type of analysis can be integrated into dashboards, scorecards, or business reports. Descriptive data helps in problem identification and classification. Predictive analytics uses historical data to train projection and forecasting models to determine potential future outcomes. Prescriptive analytics use data to help decision-makers make more informed strategic decisions. It answers the question of what should be done, instead of what have we done, or what might we do in the future (Sharda, 2020, pp. 31–32). Some analysts also include a fourth category called diagnostic analytics for analysis of root cause answering why an event occurred.

The classic tool used for populating and conducting analysis on data is the spreadsheet. Spreadsheets with add-ins can be used to conduct decision modeling using approaches such as linear programming methods, making the process accessible to those without extensive background experience with data modeling (Sharda et al., 2020, pp. 474-477). Decision situations that are simple can be modeled using decision tables, but complex issues require the use of decision trees or simulation. Simulation is important to business intelligence because it find characteristics related to solutions alternative than the predictive outcome (Sharda et al., 2020, p. 493).

Enormous, rapid leaps forward in development of accessible business intelligence tools can be attributed to the hard work and diligence of open-source communities. Python is a versatile programming language favorited in the data science industry. Packages are used to design models for practically any statistical approach. Package development may even include multidisciplinary experts to design packages exploring revolutionary modeling processes. Spiking neural network (SNN) simulators replicate neural activity to develop rapid prototypes and data models throttled only by the Von Neumann bottleneck (Zou et al., 2021).

Artificial intelligence (AI) revolution is becoming an increasingly enormous facet of business intelligence. AI is changing how we interact with our world through innovations in technologies like chatbots, knowledge systems, robotics, machine vision, neuromorphic technologies, and natural language processing. Automatization of processes enables business to refocus their resources to increase efficiency and profitability.

The relationship between data and visualization is an interesting aspect of business intelligence. As technologies capture data more effectively than ever before, the struggle to turn the numbers into meaningful information has become more important than ever before. After information is gathered and evaluated, it must be transmitted to board members, managers, staff, and other stakeholders. Skilled communicators are capitalizing on age-old storytelling skills to guide an audience through data analysis. This brilliant strategy includes developing characters, plights, obstacles, and lessons. Story-boarding the various elements clarifies the underlying information being conveyed (Sharda et al., 2020, p. 179).

Business Intelligence Data Quality

Quality of data is an important consideration before exploring data models best suited for a business intelligence application. Data quality is characterized by:

· Source reliability

· Content accuracy

· Accessibility

· Security/privacy

· Richness

· Consistency

· Currency

· Granularity

· Relevancy

Some quality characteristics can be corrected by preprocessing, where data are cleaned and transformed prior to analysis (Sharda, 2020, pp. 125–132) Preprocessing steps include consolidation, the process by which data are collected and consolidated. Cleansing is then conducted to impute missing data and reduce noise from outliers. Transformation processes normalize and aggregate data, while reduction and narrow the set down to relevant features (Sharda, 2020, p. 132).

Business Intelligence Problem Analysis and Modeling

The first step to modeling a business intelligence solution is the intelligence phase, where objectives are determined, data sources are examined, and data is collected. Problems and opportunities are identified, classified, and assigned ownership. Output from these activities is a fully developed problem statement. Designing a model involves evaluating requirements to determine criteria. Models are structured based on business objectives, but alternatives are documented. The model is then tested through predicting outcomes, then measuring performance to determine validity of the model. Activities in this phase result in a validated model. Unlike traditional life cycle development processes, business intelligence modeling requires close examination of alternatives. As business requirements and strategic ambitions develop, alternate validated models may provide better insight for decision makers. The choice phase also provides analysts with an opportunity to experiment with alternative models to ensure the final selection is the proper solution. Sensitivity and what-if analysis are performed on all models, then compared to determine which performs best. This process is visualized in .

Figure 9.

Figure 9

BI Modeling Phases

Business Intelligence Ethics

Ethical considerations for widespread usage of business intelligence strategies is a hot topic in all industries. Accessibility of software and applications has, in many ways, leveled the playing field for small and medium sized businesses. However, there is a common misconception about the relationship between human input and intelligence system output. Analysis and models are often incorrectly regarded as infallible, causing enormous societal harm.

A study by Twitter showed amplification of voices within political factions world-wide through social media prioritization algorithms used on their platforms (Huszár et al., 2022). Another study evaluating algorithmic bias in human resources found models trained using historical data falsely correlates male gender with managerial capacity (Vassilopoulou et al., 2022). Most harmful is the bias in healthcare intelligence applications. Racial bias in algorithms assessing healthcare costs concluded black patients are less sick than white counterparts erroneously because training data had far more instances of white patients to evaluate (Norori et al., 2021). As with any other technology, the correct approach to any business intelligence solution should be “garbage in, garbage out.”