Chapter 9 * Decision Making and Information Systems * Bob Travica Chapter 9Decision Making and Information SystemsDecision making is one of essential management tasks. Effective decision making is informeddecision making. Managers get informed via information systems, oral communication, andpossibly in other ways. This chapter explores decision making from the perspective of astandard rational model and two alternatives that exist in reality. The chapter also discussesinformation systems for making decisions at different levels of management – DecisionsSupport Systems (DSS), and TPS and MIS, which were already described in great detail.Decision Making and ManagementA big part of management is decision making. It is involved in almost anything managers do. Aclassical list of managerial tasks includes planning, organizing, staffing, delegating or directing,coordinating or controlling, reporting, and budgeting (note the acronym POSDCORB). Some ofthese tasks are a direct application of decision making, such as planning and delegating ordirecting. Other tasks usually result in decisions. So for example, organizing work inorganizational departments and offices requires analyzing a current work situation and the nextstep may be deciding on changes. Similarly, hiring new employees and assigning employees tojobs (staffing tasks) also end up with a management decision.A decision is about choice making. A decision maker needs to have two or more choices(options) available and then choose (select) one of those that makes the decision. Recall that inprocess diagrams a decision is represented with a question inserted in a diamond shape,followed by optional output steps resulting from possible answers (choices). In more detaileddiagrams, the decision diamond can be broken down to an entire process.The choices can be carefully evaluated to arrive at the best one. This is the case of ideal,rational decision making. However, when decision makers work under some pressure they mayneed to settle for a choice that is good enough rather than perfect. Even more of deviationfrom rational decision making happens when decision making is performed over prolongedperiods without delivering a clear decision.Any decision is made for a purpose. When a manager faces some problem, she/he concentrateson it in order to find a solution. As there is a start point (a problem) and the end point (adecision), there must be some activities in between these. Altogether, they make a process.Here are some examples of the problem that can initiate a decision process: supplies are being1

Chapter 9 * Decision Making and Information Systems * Bob Travica expended more quickly than planned, a job position gets vacant and has to be filled, a budgetmust be allocated between purchase requests, annual bonuses are to be awarded, strategicgoals need to be defined, and so on. Decisions can apparently apply to everyday operations(e.g., delegating tasks to subordinates), a close future (e.g., monthly purchases for replenishingthe inventory), and a more distant future (strategic goals setting).Once a decision is made, a decision maker needs to ensure that it will really solve the problemit was made for. This includes additional steps of monitoring decision effects and of adjustingthe decision if the effects are not as expected. Only when a decision really solves the problem,the problem solving process is over.Decision making processes are data-intensive. A manager may need various reports, businessdocuments, analyses, and direct communication in order to get prepared for making effectivedecisions. The scope of data coverage depends on the level of management and the problemdealt with. In addition, decision making requires knowledge. In particular, knowledge ofbusiness is a part of management competence. This knowledge is practical experience ratherthan theoretical knowledge, and it facilitates effective informing of the manager. All IS typessupporting management, which were mentioned before, assist in decision making.Process of Rational Decision Making and Problem SolvingHerbert Simon developed a rational model of decision making and problem solving, whichintended to raise this management task to the level of science. Figure 1 depicts this model as acircular process. The process starts with the problem identification. Competent managers arefamiliar with the workings of their organization and have a sense of potentially problematicareas. Still, they need to do their homework and learn about a problematic situation as soon asthe first signals occur. At this point, TPS and MIS play an important role. Managers use reportsand queries from these systems in order to recognize potential problems requiringmanagement attention.MIS exception reports are particularly helpful at the problem identification step because theseare automatically created when a significant deviation from a planned organizationalperformance occurs. For example, a processed food producer may have a sudden drop in salesin the past quarter. This drop would trigger a purchasing MIS to create an exception report. Thisreport could be sufficient for a sales manager to identify the drop in sales as a problem. TheMIS exception report also helps in identifying the products with the declined demand and theirbuyer (let us assume it is one single distributor company).Once a sales drop is detected as a problem, the manager-decision maker has to dig deeper andnarrow down the problem definition. The manager gets into a role of investigator who makeshypotheses about the source of troubling sales, examines them, and filters them out untilreaching the most valid definition of problem. But before going any further, the manager wantsto check historical figures in order to see whether a drop in sales of the problem products has2

Chapter 9 * Decision Making and Information Systems * Bob Travica ever occurred in the past. The exception report may or may not show such historical figures. Ifnot, the MIS usually has such querying capability. It usually provides the answer in graphicalformat for quick informing. Finding out that the current drop in sales is cyclical may actually bethe end of the decision making process, as there is really nothing that can be solved in thatcase. The sales will rebound anyway, once the period of the cyclical drop is over.But if the sales drop is not a cyclical event, the manager turns to the leads from the MISexception report – the underselling products and their distributor. He may want to explore ifthere is anything in the characteristics of these products that can define the problem in morespecific terms. For example, what are the ingredients of those food items? An ad-hoc query in aproduct database (which is a part of a production TPS) would answer the manager’s question.He finds that the products contain higher concentrations of fats and sugars. Does that fact turnsthe consumers away? It might, but the manager does not really know for certain and needshelp of market research.Let assume that this company does not have the capability of researching consumer markets,so help must be sought outside. At this point, the decision process is escalated to highermanagement levels, as the environment scanning enters in the decision process. Vicepresidents for marketing and operations joins the decision making team. A corporate businessanalyst also gets involved. Luckily, the business analyst quickly locates a market research firmthat had published reports on trends in the processed food demand within the geographicalmarkets of interest. With this help, the managerial team gets to learn that the productscomparable to their company’s underselling products have not really experienced a significantdrop in sales. Therefore, they conclude that the market does not reject the sort of foodproducts that the company sells.Figure 1. Rational decision making3

Chapter 9 * Decision Making and Information Systems * Bob Travica Rational decision making implies that all possible angles are examined at the problemidentification step until the problem gets defined perfectly. To do so, the decision making teamneeds to expand investigation inside the company to potentially relevant issues, such asproduction quality. The investigation outside the company would also need to expand byfocusing on substitute products. To keep it simple, let us assume that this investigation ends byreturning to the initial lead. The management team defines the problem as “distributor-relatedsales drop.” They are reassured of this when they discover through their sales MIS that thesales of other products to the same distributor have also fluctuated significantly in the past.With this conclusion reached, the first step of rational decision making is completed.The second step in rational decision making is about defining optional solutions. In our example,one solution may be to inquire with the distributor about reasons for the sales drop and torenegotiate terms of trade, so that the distributor accepts purchasing quotas. The next optionmay use the similar idea but defined in a more formal way – changing the contract with thedistributor with the formally defined liability for the distributor’s underperformance, andperhaps with rewards for an outstanding performance. Yet another solution may be to switchto a new distributor. This option would include searching for a comprehensible list ofdistributors operating in different geographical markets. Still another option is to bypass thedistributor and to sell directly to retailers. Furthermore, some of these options may becombined (e.g., renegotiating the terms of trade with the distributor and exploring the retailoption at the same time). Still, this is not an exhaustive option list. The rational decision makingassumes that all valid solutions are to be defined because only in that way the final solution isreally the best option.The third step in rational decision making is the evaluation of optional solutions. Themanagement team evaluates pros and cons, or benefits and costs of the options. Most of theanalysis is financial. For example, the first option in our example (“inquire and renegotiateterms of trade”) may be cheap, but its benefits in terms of increased sales are questionable.Changing the contract with the distributor may cost more in time and money (tangible costs)and it could be hard for the distributor to accept (which is an intangible cost). However, thisoption could give the company more control over sales, that is, larger benefits. The distributorswitching option could be even costlier but more capable of resolving the problem. And so on.Larger benefits cost more.In weighing costs and benefits, our decision making team puts at work the company’s financialDSS. Their system has a data modeling module that consists of formulas for processingnumerical data. For a moment, think of spreadsheet software like Excel (see more below). Afinancial analyst has entered all the solutions in the DSS. Upon a discussion, the decision makersdetermine benefits and costs to each of the solutions and make quantitative assessments ofthese. The benefits are assessed in terms of the sales increase in certain periods of time.The DSS is further instructed to weigh costs and benefits for each solution, so that it createsseveral scenarios. Assuming that costs are constant, one scenario increases benefits whileanother decreases benefits for the same percentage. Another set of scenarios keeps benefits4

Chapter 9 * Decision Making and Information Systems * Bob Travica constant, while varying costs up and down. The DSS is instructed to evaluate the optionalsolutions and to rank order them according to the benefit/cost ratio. The system delivers therequested scenarios within seconds. Now it is up to the decision makers to examine eachscenario and to select the winner. Optionally, they may need to adjust some numbers and torun the DSS again before arriving at the commonly agreed outcome.Figure 2 depicts the rational decision making process with accompanying support of MIS, TPS,and DSS.Figure 2. Rational decision making with support of ISCompleting Problem SolvingAlthough a decision is made, the larger problem solving is not over yet. A decision is officiallyput in force so that it becomes a guide for conduct. The manager-decision maker then needs tomonitor effects of the decision. He expects that the decision will resolve the problem for whichit was made. This may happen so, or not quite as expected. This monitoring makes the fourthstep in the entire problem solving process. If the decision effects are as expected, the decisionmaker does not need to intervene. However, if the decision has not produced expected effects,the decision maker has to take another step toward adjusting the decision. In the givenexample, if the sales do not recover within an expected period, the decision makers need toadjust the decision (the solution). For example, if option one was the winner (“inquire with thedistributor and renegotiate terms of trade”), the decision makers need to get back to thenegotiation table with the distributor. After adjusting the solution, monitoring of real effectsresumes.The rational model of decision making and problem solving is based on several assumptions.First, a decision maker is perfectly informed when defining a problem, creating optimalsolutions, and when evaluating them (steps 1-3 in Figure 1). Second, the model does not5

Chapter 9 * Decision Making and Information Systems * Bob Travica account for constraints, such as time and resources (human, material). While you should deploythe rational model whenever possible, you should know that its assumptions are rarelymatched in reality. This fact limits the model’s applicability. For example, it is used in situationsthat are more familiar, such as in IT purchases. When a particular piece of software or hardwareis to be purchased, you can identify comparable products and then rate them on capabilitiesand cost items. The result is a long scoring table that will clearly identify the winning product.Therefore, the rational model will certainly be valuable. Still, keep in mind that its fullimplementation requires that the list of competing products is exhaustive, the list of featuresand cost items is complete, and that a scoring table should be created for each piece of IT whileplanning an IS. Such a tall order is apparently hard to meet, and some compromising isnecessary. Therefore, even with this simple problem, the rational model may not apply in full.In real-world organizations there are even more extreme deviations from the rational model, asthe further discussion will show.Decision Support SystemDSS is deployed when important decision about an organization’s future have to be made. If thefood producer in the discussed example does not put a halt to the dropping sales of certainproducts, the company’s revenues may suffer on the long run. DSS serves higher managementlevels. In the food producer case, the problem was initially addressed by a mid-level manager,but then it was escalated up the hierarchy. The team of decision makers then used a financialDSS in evaluating optional solutions to the problem of underselling products.Another DSS that was implied in the case is a marketing DSS that the company did not own andtherefore had to rely on services of a marketing research firm. Producers that are directlyinvolved with consumer markets may own such a DSS. As the larger segment of data in thissystem comes from the organizational environment, it brings the challenge of feeding it withcurrent, complete and accurate market data. Not every company can afford responding to thischallenge.Problems to solve with a DSS are less structured, that is, less possible to understand andanalyze than those pushed down to MIS and TPS. This lack of structure is due to viewing alarger picture of an enterprise. In addition, a whole new segment of the organizationalenvironment is added. Therefore, sources of DSS data are both within a company and outsideof it. As for the organizational sources, a DSS delivers key performance indicators. Examples arethe state of cash flow in the whole enterprise, year-to-date summaries, year-to-datebreakdowns of earnings, expenditures, business hours, purchases and sales, and similaraggregated figures.The key performance indicators are usually represented visually in easy to understand andattractive formats. One such format is the dashboard represented in Figure 3. It resembles thedashboard in cars, with gauges indicating safety (the green zone of the round gauge), danger(the red zone), and neutrality (the yellow zone).6

Chapter 9 * Decision Making and Information Systems * Bob Travica Figure 3. A DSS dashboardAny DSS also has a drill-down capability that allows for investigating what is behind theaggregate figures. The user just needs to click on a particular number or button on the screento get more specific data.The content of environmental data varies with the DSS domain. Examples are competitionfigures, government regulations, product development trends, technology trends, and marketanalyses. Environmental data can also be displayed via dashboards.A model of DSS is depicted in Figure 4. It shows that DSS performs modeling and mining of data.These capabilities are supported by a core module of a DSS. As any other information system,the DSS also has the user interface (as a dashboard cited above), and some databases (e.g., forenvironmental data). The figure also indicates a link between the DSS and MIS. Drawing on TPSdatabases, a MIS outputs feed into a DSS. (Recall the concept of hierarchy of informationsystems supporting different levels of management.)Figure 4. Basic design of DSSModel-Driven DSSThere are two main types of DSS: the model-driven DSS and the data-driven DSS. The DSSdiscussed in this chapter’s case is model–driven. A model-driven DSS is a system that has a7

Chapter 9 * Decision Making and Information Systems * Bob Travica special module for analyzing quantitative data in order to get answers to particular questions.This module can perform what-if analysis, statistical tests, process simulation, and some otherkinds of analysis.A model-driven DSS for what-if analysis can modify or create input data in order to arrive at adesired result of calculations that the user programs into the system. For example, a userwishes to optimize returns from an investment given the certain input parameters (cost,interest rate, return period). The system can calculate one of the input parameters (as in Excel’sGaol-Seek) or several (as with the Solver that runs multiple structural equations). The analysiscan also move in the opposite direction, where the target result keeps changing with changes ininput parameters (in Excel, this is called scenario analysis).Another kind of the modeling core of DSS performs statistical testing. Usually, a hypothesisabout a causal relationship is tested for its acceptance or rejection. An executive may want tosee if the productivity increase is due to job satisfaction or investments in new IS, or perhapsboth of these factors combined. A DSS supporting statistical regression analysis may suggest ananswer.In contrast to the above DSS modules that work with static data, a special kind of DSS modelingcore can run simulations. An example is simulating complex manufacturing operations in theglobal space, where supplies come from distant places, and there could be unforeseen events,such as the equipment downtime, disruptions in operations and fluctuations in orders. Anotherexample is simulating business processes to test their different designs and effects on processperformance. Figure 5 is a screen shot from one such business process simulator, which can beused in a DSS. Simulations are used when decisions are made about dynamic reality, wherechanges of different aspects coincide and some values cannot be predicted with certainty.Figure 5. Business process simulator as part of DSSData-Driven DSSData-driven DSS deploy large repositories of organizational and environmental data in order tofind out new relationships and patterns. One widely used data repository is called datawarehouse. Think of a warehouse of material products, which may store various kinds of things8

Chapter 9 * Decision Making and Information Systems * Bob Travica in large quantities. The same idea applies to a data warehouse. Its data come from varioussources, including relational databases used in TPS and MIS. However, the table structureusually must be abandoned in order to integrate the data from relational databases with otherdata.Once a data warehouse is created, software for data mining is applied on it (see Figure 4). As ifthe soil is mined for precious metals, different data mining software may detect differentthings. Put in the perspective of decision making, data mining defines the problem of decisionmaking in the form of three questions:(a) Which events do flow in sequence?(b) Which events coincide?(c) Which entities go together?As an example of question A, the specific problem to solve is: Do customers buy a new TV and aDVD player sequentially within a predictable period of time? A “yes” answer may shorten thedecision making process immensely. It leads directly to management action of promoting theproducts whose purchases are related in time.An example of question B is the following problem: What do customers buy together? Forexample, it was found through mining sales data of an American grocery chain that beer andbaby diapers were purchased together by a particular profile of customer (a male of the usualparenting age) always at a particular day and time. This sort of finding can again lead toimmediate management action. In this case, the identified products can be displayed togetherin order to increase sales.An example of question C refers to identifying customer groups based on income level, age,home location, purchase period, etc. In other words, the data-driven DSS is used for marketsegmentation, as mentioned above.Note that the data-driven DSS modifies the classical rational model of decision making: if theabove questions could be taken as the problem definition, the subsequent steps in the modelare performed by the system in a straightforward manner without options characterizing therational model. Current efforts in the data-driven DSS are focused on using Big Data in additionto the traditional sources. Big Data come in forms that are more varied than the alpha-numericdata fitting relational databases.Figure 6. Sometimes data mining delivers surprises9

Chapter 9 * Decision Making and Information Systems * Bob Travica Satisficing Decision MakingTrying to make decisions in the rational manner discussed at the start of the chapter, decisionmakers carefully investigate problems to arrive at the best definition, work hard on creatingoptional solutions, and weigh these carefully. However, there are various kinds of limitations inorganizations that can derail this process. Managers often lack time and resources to get to thebest possible decision. Other limitations involve opposed views and unsupportive reaction ofthe people involved. Perhaps even more upsetting for a rational decision maker is the fact thatnot everything is always clear in organizational life. Even the first process step of problemidentification may present a stumbling block that is hard to get around. Decision makers usuallyneed some time to “digest” and investigate an initial recognition of a problem, as discussedabove.The author of the rational decision making model realized these constraints, and developed analternative model. So, Simon came up with a model of “satisficing” decision making. Heinvented this word to convey the idea of decisions that are good enough, given the constraintsunder which they are made. A satisficing decision is less than satisfying, but it may be madequickly and can serve the purpose.To understand this model, imagine that you want to buy a mouse for your PC. You have onlyfive minutes to complete the purchase in an unfamiliar electronics store. The store is large, andthere is no sales assistant in sight. You run around as the clock is ticking. Finally, you find a shelfwith computer mice and browse through it. But you cannot see your favorite product. Youchange your mind: instead of looking for that perfect mouse you will go for one that will do thejob for the moment. At the least, the mouse should sit well in your hand and be on the cheaperend. You spot one cheaper mouse, and you try it. It feels OK in your hand. Here is another. A bitless comfortable but the price is right. There is no time for more searches. You decide to go forthe lower price since you will not use that mouse forever, anyway.Figure 7. Decision MakingWhat you have done in this example complies with satisficing decision making (see Figure 7).You purchased a mouse under a severe time pressure, in an unfamiliar environment, and withno help. Realizing the constraints, you simplified the problem definition (“buy an affordablemouse that fits your hand”). Then you found two candidates and ended the search. Assumingthat you are buying a throwaway thing, you stacked your final decision on the price, and pickedthe fitting product.10

Chapter 9 * Decision Making and Information Systems * Bob Travica More often than not, organizations have to make decisions under similar and even worseconstraints. There are limitations in time, organizational resources, conflicts between factionswith different interests, as well as in cognitive capabilities of decision makers. Therefore, quickand imperfect solutions are a real rescue in these circumstances. They can do an important parta job in an acceptable manner. For example, instead of running an elaborate rational decisionprocess when buying a PC, the problem can be defined so that a “good enough PC” is sought. Itshould have a large RAM and a mid-range speed of CPU, and be in a mid-range price category.An online search of PC products can deliver screens of corresponding products. In quickdecision making, you may look at the first screen or two and pick out the first PC that matchesyour criteria. The trouble with “good enough” decisions is when the temporary solutions theydeliver become permanent, and when decision makers get used to make most of decisions thatway.MIS and TPS may help in satisficing decision making by shortening the time for steps 2 and 3 inthe model shown in Figure 7. Assume that the electronics store had an online catalog on itsWebsite. You could browse it in the store or on your smart phone while making stops duringthe drive to the shop. You could set search conditions for the size and price and get the picksinstantaneously and neatly sorted. Informed so well, you would enter the store with the clearidea of what you wanted to buy.Evolutionary Decision Making and Problem SolvingAs the service sector is capturing an increasing part of economy, it is important to knowsomething about the way decisions are made in it. In particular, American public hospitals asthe place for decision making were on the research agenda of Charles Lindblom. He found thatdecision making and problem solving unfolded through a more flexible process than thosediscussed so far. However, this process was usually complex, involving too many players. Anadministrator could not make a decision without other stakeholders (internal and external). Theexact number of stakeholders was not always possible to plan in advance. Struggles overbudgets and assets were a norm rather than an exception, making the process very political.Nobody was able “to cut the knot” and move the process decisively forward. Theuncontrollable process complexity led to problems in process coordination.The sub-optimal process design resulted in suboptimal performance of the decision process inthe hospitals Lindblom investigated. Most apparently, the start-to-end time was out of controlas decision making in some cases dragged in nearly infinite loops. Some processes could noteven get past the first step of defining the problem. Decision makers would engage inprolonged negotiations and maneuvering in order to protect their interests. Worse yet, decisionmakers would reopen the solution repository after one solution was already put in effect. Theconsequence was a prolonged decision process time. All in all, decisions evolved via slow,complex processes, rather than being made in a rational fashion. To characterize this, Lindblomused metaphors of muddling through and zig-zig movement.11

Chapter 9 * Decision Making and Information Systems * Bob Travica Figure 8 represents some of the characteristics of evolutionary decision making. Solutionoptions are defined tentatively because decision makers cannot agree on clear-cut definitions.An optional solution is implemented just partly. If there is a strong push back, the administratorswitches to an alternative. This may happen as many times as the blockages occur, resulting in azig-zag path. And it may so happen that after trying many optional solutions, decision makerseven get back to the first option that was discarded a long time ago.Figure 8. Evolutionary decision making processEvolutionary decision making may deploy TPS, MIS and even DSS. However, given themaneuvering and struggling aspects of this process, it is difficult to predict if the systems willreally be used and what impacts they can make. Nevertheless, enforcing the use of IS in publicorganizations may contribute to more effective channeling of their decision processes.12

Chapter 9 * Decision Making and Information Systems * Bob Travica Questions for Review and Study1. Define steps in the rational decision making. (Simon’s model)2. What is the relationship between rational decision making and problem solving? (Simon’smodel)3. How do information system

In weighing costs and benefits, our decision making team puts at work the company’s financial DSS. Their system has a data modeling module that consists of formulas for processing numerical data. For a moment, think of spreadsheet software like Excel (see more below). A fi