User-Centered Business Intelligence

Published on May 2016 | Categories: Types, School Work | Downloads: 24 | Comments: 0 | Views: 134
of 10
Download PDF   Embed   Report

In today’s volatile economic climate, it has become very difficult to predict and forecast business activities. Rapid and ad hoc changes in financial markets, tighter regulations, and fluctuating consumer behavior have all made it challenging for organizations to understand with certainty what is happening within their businesses.

Comments

Content

USER-CENTERED BI

User-Centered Business Intelligence
Patrick Foody Abstract
In today’s volatile economic climate, it has become very difficult to predict and forecast business activities. Rapid and ad hoc changes in financial markets, tighter regulations, and fluctuating consumer behavior have all made it challenging for organizations to understand with certainty what is happening within their businesses. Patrick Foody is technical director at Neutrino Concepts Limited. [email protected] Traditional business intelligence (BI) tools, designed to help organizations understand their activities, appear to offer little insight. Many systems require complex queries that can be written only by experts, and information is often delivered several days or weeks after it was first requested. This article will look at a new approach to BI systems: user-centered BI. It will discuss the need for the BI industry to turn its approach inside out and start with the user rather than being bound and driven by what the technology can do. Both the business case and the user case for user-centered BI will be discussed, as will the approach to user-centered BI system design.

The State of the (Business Intelligence) Nation
In today’s volatile economic climate, it has become very difficult to predict and forecast business activities. Rapid and ad hoc changes in financial markets, tighter regulations, and fluctuating consumer behavior have all made it challenging for an organization to outline with certainty where its business is going next. Coupled with the data explosion, businesses don’t know what data they should be analyzing, what questions they should be asking of that data, and, in some cases, how to extract what they need to know. To date, we have huge and vast warehouses of information (terabytes or petabytes). This, coupled with an

BUSINESS INTELLIGENCE JOURNAL • VOL. 14, NO. 4

17

USER-CENTERED BI

ever-widening variety and choice of external information from the Internet, means business analysts are literally swamped with data. The paradox is that business analysts are drowning in a sea of data but unable to obtain the knowledge they need to address the more difficult business problems.

What is the Answer?
For BI systems to truly deliver their intended value, further advancement in systems design is needed to complement user processing needs. The goal of usercentered BI is to focus on the true essence of business intelligence—important, meaningful, and actionable information delivered in a timely manner. The most powerful resources for tapping into BI’s value are those that engage the tremendous capacities of human perception, human cognitive reasoning, and human intelligence. In short, user-centered BI applications need to combine the strengths of self-service BI technology advancements with computer-aided cognition to support problemcentered decomposition of business problems. User-centered BI applications also require a new approach to data understanding, collection, indexing, and access—one that is faster, multifunctional, dynamic, and easier to use.

To date, the majority of BI systems have been designed in light of how technology approaches a problem or from the understanding that an analyst would have of the system.
This has caused what has been termed “the dilemma of data overload,” where many organizations aren’t effectively analyzing the data they do have to improve their businesses. What is more troubling, perhaps, is that many companies that purchase powerful analysis and business intelligence tools don’t use them effectively. What is a business to do? Before jumping headfirst into yet another large-scale BI system investment, it is worth noting that Gartner put self-service BI on its list of top 10 strategic technologies for 2009. However, if self-service BI is to prove as groundbreaking as it promises, it must be designed to truly facilitate user-centered analysis. To date, the majority of BI systems have been designed in light of how technology approaches a problem or from the understanding that an analyst would have of the system. As a result, the technology, which is intended to be easy to use and aid human cognition of a business scenario, interferes and confuses rather than helps the user or clarifies a position. User-centered BI is the answer to today’s BI shortcomings. We will examine the business and user case for it and the approach to user-centered BI system design.

The Business Case for User-Centered BI
Clearly, providing more “relevant” and “action-oriented” information is a pivotal issue for C-level executives as they step up to a more strategic leadership role at the center of management for the entire organization, especially during tough economic times. One of the keys to improving the quality and timeliness of information is to implement a holistic, user-centered analytical framework that is designed to enhance decision-making across the entire value chain and at all levels of the organization. By doing so, management can provide tremendous tangible value throughout the company. This is the design philosophy underpinning user-centered BI applications. The technology provides a unique solution to the management of business problems, providing real-time exploration for decision-makers as well as for departmental clusters of information users. It is unique in its approach to decision-making definition and execution and remains close to standards initiatives. The right way to address complexity and minimize risk is by adopting an interactive, user-centered, analysis-

18

BUSINESS INTELLIGENCE JOURNAL • VOL. 14, NO. 4

USER-CENTERED BI

based approach, focusing on operational procedures that provide an effective strategy for managing business complexity and scale. Decision-makers who embrace a successful blueprint for a user-centered analysis approach will derive the following business benefits:
n

Executives will be able to align the goals and related measures of the individual business units and functions with the strategies of the organization. Further, it will help identify and rationalize measures that may actually conflict across business units. The organization will be able to quickly sense and respond to change, both internal and external. It will be able to do this by aligning and integrating people, technology, and information with the core business processes across its entire value chain. Analysts will be able to dynamically structure, format, and prepare data in near real time, reducing the cost and effort of data mining initiatives. Furthermore, the application will deliver the results of such analysis to the appropriate decision-makers when it is required so that companies can analyze events as they occur. Some refer to this concept as “zero-latency” or “real-time” decision-making. The traditional BI approach has meant that analysts have had to wait until data is extracted, formatted, and prepared for analysis, by which time it is usually too late to take corrective action. It will enable role-based information delivery to decision-makers and other information consumers.

This would be a healthy situation if the pivot table were an effective and easy-to-use interface. However, research by Peter O’Donnell, a lecturer in the Decision Support Systems Laboratory at Monash University in Melbourne, Australia, established that using a pivot table to search for data in an OLAP database requires a lot of work (O’Donnell, 2005). The starting point for analysis using an OLAP application might make sense for one user but could be confusing for another. Users of these products often generate the most basic and obvious reports and never achieve the deep analysis that more complex business problems need. Gartner corroborates this research, citing that the lack of both end user and developer skills is frequently a major barrier when deploying BI applications. According to Gartner, “Anecdotal evidence suggests no more than 20% of users in most organizations use reporting, ad hoc query, and online analytical processing (OLAP) tools on a regular basis” (Schlegel, 2008).

n

n

The starting point for analysis using an OLAP application might make sense for one user but could be confusing for another.
Further, to understand the data being viewed, users must understand the complex, multidimensional data structure they are manipulating. Most OLAP data structures are complex. For example, the standard demonstration Adventureworks Internet sales data structure, shipped by Microsoft with the Analysis Services product and used in many books and tutorials on OLAP, has 14 dimensions and two fact tables. It is not unusual for production OLAP databases to have an even larger number of dimensions. The hierarchies within individual dimensions might also be a source of confusion or frustration for users. For example, a search for a specific product in a typical

n

The User Case for User-Centered BI
BI is often described as a technology that helps users visualize their business. The interface provided to users needs to aid that visualization and assist in their exploration of data. It must be easy to understand, simple to use, and relatively intuitive. Currently, there is little to distinguish the combination of graph and pivot-table-style interfaces offered by the core products of the major vendors in the BI client market.

BUSINESS INTELLIGENCE JOURNAL • VOL. 14, NO. 4

19

USER-CENTERED BI

Metadata tagging

Accept and format data Problem-centered decomposition and source analysis
Data

Hierarchical decomposition

Accumulate, Fuse, and Discover Information

Data filtering, correlation, and fusion
Monitor Process and User; Adapt for Improved Inferences

Formulate and refine queries

Format and display
Interaction and collaboration with other human and virtual agents

Reports

Evaluate hypothesis
Assess and Analyze Hypothesis

Formulate and refine alternate hypothesis

Decompose problem

Accumulate, filter, and fuse data Retrieve information
Figure 1. User-centered analysis process

product dimension—where there could be tens of thousands of fields—might require a user to remember specific information about where a product has been located in a multilevel hierarchy of product catalogs, groups, categories, and lines in order to find the specific item she is looking for.

What is the User-Centered Analysis Process?
To improve user-centered information processing, the analyst is considered to be the center of an ongoing, emergent, and evolutionary process that accesses enormous amounts (petabytes) of collected data to recognize a problem of interest. Figure 1 shows a single analyst at the center of this process. However, it is recognized that a team of analysts may construct this process and that these analysts may not in fact be physically co-located. The top part of the processing cycle in Figure 1 involves the collection, accumulation, and fusion of data. The process is shown as a cyclical process that involves seven steps. Note that these steps are illustrated as being sequential. However, in an actual system, the steps could be integrated. In addition, these functions may be automatically performed by processing functions (for example, scheduled or daemon processes), by human analysts, or by a hybrid human-computer analysis process.

A New Approach
Analysts need applications that are user-driven, not application-driven. They need simpler interfaces to allow users to explore different hypotheses, depending on the focus or direction of their business questions. Unless this happens, the use of OLAP technology in business intelligence systems will be restricted to power users who have the skill and confidence to master a complex interface. As a result, OLAP-based systems may not contribute to improvements in the design of business intelligence systems or improve corporate decision-making. Effectively, the BI industry needs to turn its approach inside out and start with the user rather than being bound by what the technology can do.

20

BUSINESS INTELLIGENCE JOURNAL • VOL. 14, NO. 4

USER-CENTERED BI

Data Acquisition and Formatting of Data Incoming and collected data is monitored, coordinated, accepted, and formatted by the data acquisition processing system of the user-centered application. Here, it is assumed that data includes centrally controlled information platforms. These could include enterprise data warehouses, locally controlled information sources such as local data marts (maybe even Excel), or external information resources such as the Internet. The data formatting in this process should not be confused with formatting in the traditional extract, transform, and load (ETL) mechanisms of the data warehouse. The formatting here could involve the analyst importing data from a local spreadsheet file on the user’s desktop and then applying Excel algorithms to transform and transpose the data locally. Finally, the data could be inserted back into the centralized information architecture for further, ongoing analysis. This information could then be made available to other users, who could access the finished result without having to understand the previous steps. Metadata Tagging and Transformations Data is tagged with explanatory information to augment it. Examples include the use of a prespecified domain to assist in characterizing the information, extracting key words and descriptors, and annotating the data with parametric information related to location, identity, or characteristics. Currently, these types of tagging are usually performed only on textual information. Advanced processing is performed to automatically characterize information (for example, to recognize characteristics of the data at the point of capture). This development of data tags provides the basis for rapid data retrieval, association, correlation, and fusion. Hierarchical Decomposition and Pattern Recognition This step involves the decomposition of the data into components. For operational data, this may entail entity segmentation and hierarchical decomposition into smaller components. The application must enable the user to group data elements in different ways so that different analysis can be performed. For example, an organization hierarchy can be reorganized, creating a new division

and assigning existing departments to that division. This hierarchy can now be used for further analysis, whereby future performance metrics are measured based on how each part of the organization performed before and after a reorganization. This is known as “as-is, as-was” analysis. Data Filtering, Correlation, and Fusion Data may be filtered, correlated, and possibly fused. Examples of correlation include identifying pieces of information that relate to an event, situation, or specific area of interest. A component of this function may include link analysis to identify and link related concepts or semantic “quanta” of information, or it could be data from different systems with the same characteristics. For example, the count (quantity) of Item 1 from System 1 over a 12-month period can be visually merged onto the same chart object as the count of Item 2 from System 2 over a 12-month period. Even though the data has come from different systems with different atomic-level structures, the data acquisition phase has abstracted (via aggregation) each data element to a level at which all elements can be compared on the same graph. Format and Display Information analysts need tools that make it easy to shift from one perspective to another while exploring and analyzing data so users are encouraged to pursue every question that arises during the process. If they are not distracted by the mechanics of using the software or forced to go through time-consuming steps to get from one view to another, users can become immersed in the data and the analytical process. They’ll be able to spend their time thinking about the data, not struggling with the software. According to Stephen Few’s book Show Me the Numbers, this is an underwhelming area of existing data warehouse technologies, tools, and techniques. Visual Query Formulation and Refinement With static charting, the user creates a data set with a query tool, such as a report writer, spreadsheet, or database, and then charts the data set to create visualization. To obtain a different visualization, the user must return to the query tool, create another data set, and chart the new data set to create a visualization. Visual Query greatly streamlines this process, without requiring a separate

BUSINESS INTELLIGENCE JOURNAL • VOL. 14, NO. 4

21

USER-CENTERED BI

query tool or knowledge of query languages. When the user selects a portion of an active visualization, that portion becomes a new data set, automatically charted and ready for the next iteration. With a bar chart of customers sorted from least profitable to most profitable, a user might select the short bars (to study why some customers are not profitable), the tall bars (to identify profitable customers who should be retained), or midsize bars (to be sure the needs of the average customer are met).

could conceptually be other human agents, software models, or mashups of Web services.

Access and Analyze User-Centered Hypothesis
The assessment and analysis process is shown in the lower part of Figure 1. Conceptually, an analyst will be tasked (or self-tasked) to analyze an evolving situation, strategy, or competitive threat. The analyst has access to an enormous amount of information via the continuing data acquisition and distribution process of the upper cycle. However, the analyst is not interested solely in acquiring massive amounts of data. Instead, the analyst seeks to develop hypotheses about an evolving situation and to assess and analyze these hypotheses. The process for analyzing these hypotheses is illustrated in the cyclical process at the bottom of the diagram. As with our previous discussion, these steps are not necessarily performed in sequence and, as with the upper part of the process cycle, these steps may be performed by a human, an automated computing process, or by a hybrid human-computer operation. The process involves a number of steps, as summarized below. Formulate and Refine Alternative Hypotheses The user-centered application enables the analyst to continually formulate alternate hypotheses (or tentative explanations or interpretations) of an evolving situation. The application assists the analyst in focusing attention on possible situations and helps in defining what information has been collected and the different associations and relationships that can be analyzed. The user-centered BI application has to have an integrated search capability that will index large amounts of structured data and deliver graphical representations of information to users via a mechanism (search) that is much more intuitive than traditional ad hoc query and reporting. Decompose the Problem The decomposition of the problem from a general business question or hypothesis to specific sources of required information is the inverse of the inference hierarchy shown in Figure 1. Problem-centered decomposition involves transformation from a general business question to multiple specific business questions.

The user-centered application enables the analyst to continually formulate alternate hypotheses (or tentative explanations or interpretations) of an evolving situation.
Problem-Centered Decomposition and Source Analysis The concept of problem-centered decomposition can be used to decompose general questions into specific questions. Such decomposed types of information requests may guide the overall ingestion, accumulation, fusion, and knowledge discovery process. Central to the ongoing upper loop of accumulation and discovery is the role of analysts in assisting in the interpretation of the information and managing the accumulation and discovery process. The user-centered application enables the actions of the analyst and the ongoing process to be monitored with semiautomated adaptation to improve the effectiveness of the analysis process and the discovery process. This part of the process is depicted in Figure 1 with the analyst at the center of the diagram. In addition, the process should enable the analyst to interact with other agents or analysts. Hence, a domain specialist may interact with specialists who are knowledgeable about specific accounting systems or types of data. These external collaborators

22

BUSINESS INTELLIGENCE JOURNAL • VOL. 14, NO. 4

USER-CENTERED BI

The next generation of user-centered BI tools need to have powerful search capabilities similar to Google to live up to the demands and expectations of increasingly diverse BI users. These tools should be able to search and report on traditional data in warehouses and transactional systems, as well as new layers of metadata and external data sources. They also need to locate, tag, and index all that data as quickly as Google search engines do with Web documents. The search engine must be able to search data warehousing transactional data using simple questions, not queries. For example, the user needs to search for products by names and descriptions and not primary keys (such as inventory numbers). In addition, the search-based mechanism must enable the user to enter any combination of words (instead of the hierarchical folder navigation structure of traditional BI tools) to access the content of large data stores. Integrated search needs more than the data from a star schema’s fact table; it needs to apply search criteria to structured data sets to retrieve and explore data, even when the reports don’t exist. Keywords from a search string can be mapped to an index of structured data sources and results can be returned and filtered very quickly (thanks to the fast index). This would be a very easy and fast method of providing powerful ad hoc query analysis capabilities for data retrieval and exploration, without requiring much IT investment. Retrieve Information Given specific essential elements of information required to address the business question or hypothesis, the application assists the analyst in the retrieval of this information through the application’s transformation, fusion, information retrieval, and storage layers. For instance, most utility companies monitor grids by superimposing colors and shapes over a map to represent load, usage, and outage. Live data feeds interact with an active visualization to dynamically update time-sensitive data content in real time, such as hazard and survival curves. Accumulate, Filter, and Fuse Data In support of the evolving analysis, data and information are accumulated and filtered. Components of informa-

tion fusion are required to sort, accumulate, correlate, and fuse information that may support or refute alternative hypotheses. In this case, transformations are sought between low-level data and more general representations and inferences. This part of the process is facilitated by the tight integration of the application’s transformation, fusion, information retrieval, and storage layers.

The ability to change the IT infrastructure faster than competitors to support evolving strategy and adapt to business conditions will lead to significant competitive advantage.
User-centered BI applications should have a serviceoriented architecture (SOA) at their core. This provides the analyst with a layer of components that coalesces data and information into useful, reusable modules. This, coupled with a model-driven architecture that is based on a visual drag-and-drop development style, will make it easier to build BI applications. Self-service SOA is an architectural platform for rapidly adapting and solving business problems, and because it requires few programming skills, it drives user-centered analytical application development. The ability to change the IT infrastructure faster than competitors to support evolving strategy and adapt to business conditions will lead to significant competitive advantage. User-centered BI technologies move from tightly coupled to loosely coupled information structures to deal with inevitable change. The goal of this concept is to create components that have few dependencies between them so that they can be easily combined for different purposes with no or minimal effort. This represents the other end of the spectrum from traditional data warehouse architectures, which have many interrelated dependencies.

BUSINESS INTELLIGENCE JOURNAL • VOL. 14, NO. 4

23

USER-CENTERED BI

Evaluate Alternative Hypotheses Alternative hypotheses may be evaluated in a quantitative way to provide support or refutation of the alternatives. Simulation tools may be used to project the consequences and likelihood of the hypotheses. Evidential reasoning and course-of-action analysis can be utilized, if required, in this step. Presenting multiple chart types simultaneously helps the user make multiple interpretations. The charts must link together so that active operations such as visual queries performed on one chart are applied to the others automatically. Format and Display Finally, presenting multiple chart types simultaneously helps the user interpret data in multiple ways. The charts must be linked so that active operations (such as selecting a specific series for further analysis) on one chart are automatically applied to the other charts that the analyst has grouped together as an area of interest. For example, charting sales of multiple products across multiple geographies results in an overload of visual information. Instead of further increasing the complexity of the visualization by adding a time dimension, users can animate the chart to show one time slice after another. This reveals seasonal fluctuations of specific products per geographic region. The inferences and supporting data are formatted and displayed by the application to the analyst for evaluation, refinement, and continuation of the process. When complete (or at least suitable for reporting), the analyst may create a report or publish the findings as a template for evaluation by other analysts or decision-makers. Here again, a team of analysts, supported by external human analysts, models, and Web services, may conduct this process. Ideally, when collaborating with external resources, the analyst should not need to know (nor care) whether his collaborator is a human, a software model, or a Web service.

Summary
In basic terms, the goal of user-centered BI is to help organizations make better decisions faster. By better decisions we mean decisions that are fact-based and lead to better results for the company overall, resulting in increased shareholder value. This ranges from setting corporate strategy to day-to-day decision-making by individuals at all levels of the organization. The key to making better decisions faster is having the right information available when decision-makers need it. There are a few important dimensions to improving the quality and timeliness of decision-support information. One is broadening the type of information that is typically available to managers, moving from an over-reliance on monthly financial information to a better balance of real-time or what can be referred to as “right-time” financial and nonfinancial information. Another is focused on facilitating the cognitive processes involved in information fusion and development of computer-based cognitive techniques that can support and improve the cognitive process. Aligned information architecture is adaptive in that it has the ability to quickly adapt to internal and external changes. It does this by aligning and integrating people, technology, and information with the core business processes across the entire value chain. The user-centered business model is focused on measures of success, responsive to change, variable in its cost structures, and resilient in dynamically managing business performance. These technologies have been purposely engineered for hypothesis-based processing and ad hoc analysis (especially where large volumes of data are involved). Specifically, user-centered installations will offer orders-of-magnitude improvements in decision-making capabilities while requiring significantly fewer and lowercost IT skills or hardware resources. In other words, it provides much better performance for less money. n

24

BUSINESS INTELLIGENCE JOURNAL • VOL. 14, NO. 4

USER-CENTERED BI

References
Christensen, Clayton M. [1997]. The Innovator’s Dilemma, Harvard Business School Press. Few, Stephen [2004]. Show Me the Numbers, Analytics Press. Hostmann, Bill, and Kurt Schlegel [2006]. “Market Business Intelligence Platforms: Round Two of Consolidation Begins,” Gartner Research, ID Number: G001421, September 15. Kuperman, Gilbert G. [2001]. “Cognitive Systems Engineering for Battlespace Dominance.” http://www. dtic.mil/cgi‑bin/GetTRDoc?AD=ADA399494&Locat ion=U2&doc=GetTRDoc.pdf Llinas, J., and D. Hall [1994]. “Dual-use of Data Fusion Technology: Applying the JDL Model to Non-DoD Applications,” Fourth Annual IEEE Dual-Use Technology and Applications Conference, SUNY Institute of Technology, Rome, New York. McNeese, M. D. [2000]. “Socio-Cognitive Factors in the Acquisition and Transfer of Knowledge,” Cognition, Technology, and Work, 2; pp. 164–177. ——— [2001]. “Situated Cognition in Distributed Multi-national, Multi-cultural Teams: A Research Prospectus Informing Sociocognitive Technologies,” final report submitted to the United States Navy, SPAWAR Center, San Diego, California. ——— [2002]. “Pursuing Medical Human Factors in an Information Age: The Promise of a Living Lab Approach,” proceedings of the Human Factors, Medicine, and Technology Symposium, Baltimore, Maryland, University of Maryland Medical Center. ———, and M. Vidulich, eds., [2002]. Cognitive Systems Engineering in Military Aviation Environments: Avoiding Cogminutia Fragmentosa, Wright-Patterson Air Force Base, Ohio, HSIAC Press, pp. 230–241.

———, and M. Vidulich, eds. [2002]. “Cognitive Systems Engineering in Military Aviation Environments: Avoiding Cogminutia Fragmentosa,” Dayton, Ohio, Wright-Patterson Air Force Base, CSERIAC Press. McNeese, Michael, and David L. Hall [2003]. Proceedings of the Human Factors and Ergonomics Society 47th Annual Meeting. O’Donnell, Peter [2005].“The Problem with Pivot Tables,” Business Intelligence Journal, Volume 10, Number 1. http://www.tdwi.org/research/display.aspx?ID=7488 Rasmussen, J. [1999]. “Ecological Interface Design for Reliable Human-Machine Systems,” International Journal of Aviation Psychology, 9 (3), pp. 203–223. Schlegel, Kurt [2008]. “Emerging Technologies Will Drive Self-service BI,” Gartner Research, ID Number: G00152770, February 8. ———, et al [2008]. “Magic Quadrant for Business Intelligence Platforms, 1Q07,” Gartner Research, ID Number: G00145507, January 26. ———, et al [2006]. “BI Applications Benefit From In-Memory Technology Improvements, “ Gartner Research, ID Number: G00141540, October 2. Woods, David D. [1998]. “Designs are Hypotheses about How Artifacts Shape Cognition and Collaboration,” Ergonomics, 41, pp. 168–173. Zsambok, Caroline E., and Gary Klein, eds. [1996]. Naturalistic Decision Making, Lawrence Erlbaum.

BUSINESS INTELLIGENCE JOURNAL • VOL. 14, NO. 4

25

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close