Weitere Kapitel dieses Buchs durch Wischen aufrufen
This chapter explores the relationship between Business Intelligence (BI), analytics, and critical decisions. Section 7.1 presents a brief overview of the leading technologies relating to BI and analytics. Sections 7.2 and 7.3 examine the roles these technologies play in decision-making. We argue that BI and analytics enhance situational awareness and short-term trending rather than offer deep insights into the future. We also argue that predictive analytics run aground because of the extended durations and non-linear dynamics that typify critical decisions. As such, these technologies support operational decisions directly, but critical ones much more weakly and indirectly. Marketing fervor notwithstanding, BI and analytics hold little promise for bending the Law of Unintended Consequences on their own.
Bitte loggen Sie sich ein, um Zugang zu diesem Inhalt zu erhalten
Sie möchten Zugang zu diesem Inhalt erhalten? Dann informieren Sie sich jetzt über unsere Produkte:
The data managed by Fortune 100 companies each manage petabytes of data, hundreds of time larger than the entire United States Library of Congress. The rate of data growth in corporate data centers, for example, in industries such as financial services, manufacturing, and healthcare is expected to grow by 20+% compounded annually. Reinsel et al. [ 10]. See also Tallon [ 13] and White [ 15].
Unstructured data is often stored with tags, called metadata, such as categories or other keywords. Metadata allows massive repositories of documents (or images or videos) to be searched efficiently.
Query and reporting tools enable non-programmers to interact with databases without having to know the design (structure) of the target database tables or database programming languages.
Dashboards were popularized following the widespread adoption of Balanced Scorecard methodologies for managing businesses. Kaplan and Norton [ 8].
Silver [ 11].
Gass [ 6].
Methods include inductive algorithms and increasingly popular multi-layer neural networks. See Goodfellow and Bengio [ 7] or https://www.cio.com/article/3223191/artificial-intelligence/a-practical-guide-to-machine-learning-in-business.html for a higher level overview.
Many retailers and manufacturers automate most re-ordering using database procedures that tie into point-of-sales terminals or enterprise resource planning systems. A restocking formula might reflect the rate at which stocks are being depleted, the interval required to deliver product once ordered, the time of year (for seasonal goods), availability of qualified suppliers, and other factors.
One way to validate interpretations of situations involving high uncertainty is to project them into the future. That is, if the world is in the state that we think it is and evolves in the way that we envision, is the resulting future state plausible and logically consistent? Intelligence analysts routinely face this kind of problem in interpreting threats from hostile states and terrorists, their capabilities and intentions, and so on. They call this kind of extrapolation “hypothesis testing.”
Examples include intentional and adaptive behaviors, feedback, and delays in effects from their causes.
Taleb [ 12] makes a similar point about sensitivity to the passage of time for “fragile” systems.
Predictive analytics depend on the current situation being represented by the data set used by the algorithm to generate a pattern. If the current situation does not conform, then the patterns break for no apparent reason. Analytics work…until they don’t!.
Predictive analytics involving time series are currently limited to modeling simple temporal patterns, such as cycles and seasonal variations, not complex implementation schedules or feedback effects. And recurrent neural networks, which can “remember” and exploit sequential information (e.g., sounds in continuous speech) require massive data sets for training. Data sets for critical decisions are either too incomplete (e.g., mergers and acquisitions) or too small by orders of magnitude, to permit such applications even if algorithms were available.
The calculation of time efficiency requires calculus (or numerical approximations) to compute the area between the areas bounded by the curves for S1 and S2 and a horizontal line running along the x-axis starting at Initial Risk (R1). This amounts to treating risk like a diet: you get credit not only for taking weight off, but for keeping it off. (cf. Sect. 12.3).
Predictive analytics are similarly blind to social and psychological aspects of implementing decisions. For example, diets differ widely in how easy it is to start them, when weight loss occurs, and how easy it is to stick to them. Critical decisions are no different; options vary widely in ease of implementation, owing not only to the availability and scheduling of investment and resources, but also to social and psychological factors; managing decisions that require changes that disrupt organizational cultures and habits is the poster child for LUC (cf. Chap. 13).
You can, of course, try to address this problem by “taking more pictures.” But many decisions involve sudden changes by adaptive stakeholders or non-linear growth patterns (e.g., virality) that occur mid-stage rather than at “natural photo opportunity points” such as the transitions between stages. So there is no general solution as to when to collect the data samples needed to derive suitable predictive analytics.
All URLs Accessed 05 Jul 2019.
Ayres, Ian. 2007. Super Crunchers. Why Thinking By Numbers Is the New Way to Be Smart. New York: Bantam Books.
Biere, Mike. 2011. The New Era of Enterprise Business Intelligence. Upper Saddle River, NJ: Pearson Education.
Davenport, Thomas H. 2009. Make Better Decisions. Harvard Business Review. 87(11):117-123.
Davenport Thomas H., and Jill Dyché. 2013. Big Data in Big Companies. International Institute for Analytics. Available at https://www.sas.com/en_us/whitepapers/bigdata-bigcompanies-106461.html .
Davenport, Thomas H., and Jeanne G. Harris. 2007. Competing on Analytics: The New Science of Winning. Boston, MA: Harvard Business School Press.
Gass, Saul I. 1970. An Illustrated Guide to Linear Programming. New York: Dover Publications.
Goodfellow, Ian, and Yoshua Bengio, 2016. Deep Learning. Cambridge, MA: MIT Press.
Kaplan, R., and Norton, D., 1992. The Balanced Scorecard: Measures that Drive Performance. Harvard Business Review. 70(1): 71-79.
Jon Kolko. 2010. Sensemaking and Framing: A Theoretical Reflection on Perspective in Design Synthesis. In DRS 2010 (Proceedings of the 2010 DRS Montreal Conference). Available at http://www.drs2010.umontreal.ca/proceedings.php.
Reinsel, David, John Gantz, and John Rydning. 2017. Data Age 2025: The Digitization of the World From Edge to Core. Framingham, MA, US: International Data Corporation. Available at https://www.seagate.com/files/www-content/our-story/trends/files/Seagate-WP-DataAge2025-March-2017.pdf.
Silver, Nate. 2012. The Signal and The Noise: why so many predictions fail – but some don’t. New York: Penguin Books.
Taleb, Nassim Nicholas. 2007. The Black Swan: The Impact of the Highly Improbable. New York: Random House.
Tallon, Paul P. 2013. Corporate Governance of Big Data: Perspectives on Value, Risk, and Cost. IEEE Computer 46(6): 32-39. CrossRef
Weick, Karl E., and Kathleen M. Sutcliffe. 2001. Managing the Unexpected: Assuring High performance in an Age of Complexity. San Francisco: Jossey-Bass.
White, Tom. 2012. Hadoop: The Definitive Guide. (Third Edition) Sebastopol, CA: O'Reilly Media.
Hardesty, Larry. 2018. Study finds gender and skin-type bias in commercial artificial-intelligence systems. http://news.mit.edu/2018/study-finds-gender-skin-type-bias-artificial-intelligence-systems-0212.
Schmitt, Carolyn E. 2018. Algorithms and their unintended consequences for the poor. Harvard Law Today. https://www.today.law.harvard.edu/algorithms-and-their-unintended-consequences-for-the-poor/.
- Business Intelligence, Analytics, and Their Discontents
Richard M. Adler
- Chapter 7
Neuer Inhalt/© Stellmach, Neuer Inhalt/© Maturus, Pluta Logo/© Pluta