The business world
The growing complexity of the business world has created the increasing need for analytics; “according to a 2009 BusinessWeek Research Services survey, 83% of C-level executives agreed that the importance of using the information to run their businesses has never been greater”(Haag, Cummings, 2012). The need for analytics is coupled with a need for business intelligence and the hardware and software that comes with it. Many companies have invested large amounts of time and resources to create operational databases and data warehouses in order to perform analytics.
While operational databases support online transaction processing (OLTP) and are therefore used for day-to-day operations (inventories, purchasing, manufacturing, payroll, etc. Data warehouses support online analytical processing (OLAP) and are used for more in-depth analysis and report generating. Both are in fact supporting decision-making, just on different levels. As OLTP is aimed at routine operating decisions it is more suited to the lower levels of an organization; whereas OLAP is more suited towards the higher levels to create a long-term strategy or to deal with problems and opportunities.
Oddly enough, this matches up quite well with the two schools of thought regarding the decision-making process. The classical theories of decision-making are focused on rationality; the selection of the optimal solution is ‘guaranteed’ through the analysis of information (data) and objective evaluation of all alternatives in comparison with the objectives of the organization. [3] This theory seems more suited to OLTP, to make structured decisions; what is meant by ‘structured’ is the processing of information in a certain way to always get the right answer.
The ‘old-school’ theories work on a basis where there is some sort of agreement on the objectives between all of the stakeholders. The more recent school of thought (which can be called ‘bounded rationality) argues that stakeholders have different conflicting objectives; therefore there are many uncertain factors that can affect the decision-making process. [5] Many rational models of decision-making have been written; as structured decisions can be described more precisely and are therefore more prone to quantitative analysis these models will only be briefly mentioned in this paper.
In contrast, very few models of unstructured decisions have been created; one of them is the “trichotomy” of decision processes written by Simon in 1960; this theory is further developed in the study “The Structure of ‘Unstructured’ Decision Processes” by Mintzberg, Raisinghani and Theoret. This paper will resume the findings and the subsequent model for the ‘unstructured’ decision-making process proposed by Mintzberg et al. We will examine how rational and bounded rationality models fit in with analytics, more specifically OLTP and databases and OLAP and data warehousing.
Finally, we will see how the very book that is supposed to teach us MIS can be improved by including decision process models. In their study “The Structure of ‘Unstructured’ Decision Processes”, Mintzberg, Raisinghani, and Theoret continue the work of the other “bounded rationalists” such as Simon, going from the standpoint that strategic decisions are made up of a series of complex and unknown decisions; it is not a process one does in complete certainty, but rather mostly in uncertainty.
Using empirical evidence collected over 5 years by master students in management policy, (especially examining in-depth 25 decision processes that were deemed strategic) they have managed to create a framework for the strategic decision process. The 25 decision processes were broken down into routines and dynamic factors that then were used to create the model. Their framework is based on Simon’s trichotomy but they have replaced the terms “intelligence, design, choice and implementation” with “identification, development and selection”.
The three latter terms represent the three main phases of the unstructured decision process, which is then broken into seven routines. Simon et al. have also identified three routines that support the main framework. In addition, they have acknowledged the existence of six dynamic factors that influence the process quite significantly. The main phases are highlighted in blue and the seven routines in red. This model also takes into account the supporting routines as well as the dynamic factors, which explain the delays, the interrupts and reason the diagram seems to have a circular flow if you follow the arrows.
According to the authors, any strategic decision (therefore unstructured as there isn’t one certain answer) starts off with one or several stimuli that they categorize as a crisis, an opportunity, or a problem that evoke the need for a decision to be made; the act of doing this is the recognition routine. The decision can then move onto the evaluation/choice routine and finish there; “these two routines must be part of any decision process”(Mintzberg et al.).
They found however that all decision processes went through at least one of the other routines, and many would cycle back and forth through two or three routines. Their findings showed that all decision processes followed this general model but took different routes within it, so they created seven different route configurations each with specific characteristics of decisions. Throughout this model, there is evidence that analytics can help facilitate the decision-making process. During the recognition routine a data warehouse could be greatly beneficial for recognizing problems and opportunities previously gone unseen.
Analytics could also be used during the development phase, especially during the design stage either for modifying an already existing solution or to create a new one; it could also be used to improve the screening process, which is a very superficial routine done simply to reduce the number of different solutions into a more feasible number before presenting them to upper management, the choice/evaluation routine. The latter routine could also greatly be improved by analytics as both judgment and bargaining are made mostly by intuition.
Mintzberg et al found little evidence of the analysis routine being used. The analysis routine most definitely uses analytics in order to evaluate the possible solutions; this routine should be used more often as it seems sort of redundant to go through all these routines, to then have executives with limited knowledge of the subject in question make a decision based on their ‘feeling’. These are just a few examples of how data warehousing and analytics could be used in parallel with the model for strategic decision processes developed by Mintzberg et al.
As mentioned previously, I believe that data warehouses and OLAP are more suited to the needs of unstructured decision-making processes due to the very capable decision support systems (DSS) and data-mining tools. DSSs are developed for the very purpose of aiding in decision-making, however, there are many different tools to use; several different artificial intelligence systems, data-mining tools, and agent-based technologies. Having a model for the decision-making process at hand can help in understanding what to use when, it could serve as a guideline for using data warehouses efficiently.
My opinion is also that the framework developed for structured decisions could be used in conjunction with databases and OLTP in order to further improve the day-to-day operations decisions. We have shown that there evidently is a link between MIS and decision process models such as the one developed by Mintzberg et al. Consequently shouldn’t the very book that is supposed to teach its readers how to implement, manage and utilize information systems to optimize decision-making contain the models of structured and unstructured decision process models? The answer is yes, it should.
The book, ‘Management Information Systems for the Information Age’ by S. Haag and M. Cummings does contain brief information about Simon’s trichotomy showing the four phases of decision-making, as well as resuming his theory of ‘satisficing’; it further goes to explain the difference between structured and unstructured decisions. However, it stops there. The first issue with this, and probably the most surprising, is the fact that they chose to put the work of Simon without including any of the other research done using Simon’s basic principles such as the work by Mintzberg which develops a much more precise model that could realistically be followed by organizations in conjunction with analytics and in general as a map for using management information systems. The second issue is that they don’t explain the link between the decision models presented and management information systems, as I have attempted to do in this paper. Finally, they have made most of the information about decision-making processes very brief, only scratching the surface of the potential that is there in terms of effective management of information and information systems.
