Decision Making

Decision Making Introduction  How can you make decisions? – – – – Average the predictions of many people Meet to deliberate and discuss the decision Use the help of a person who you prefer Use some form of a price system where people who are correct are rewarded – Put the question on the internet and see how people respond 2 The Wisdom of the Crowd  The average of the crowd is often an excellent predictor. – The weight of a horse – The amount of money in a jar  It “works” when people are more likely to be right than wrong.  It does not work when people are more likely to be wrong than right (conventional wisdom is incorrect)  Surveys are good in this regard.  The problem is that there is no penalty for a wrong answer, so there is no “sorting” in a survey. You can do potentially better. 3 Committees and Discussion  The idea is to gather information in a group discussion rather than averaging. There are serious problems with this approach: – If people have similar views, group discussion tends to lead to extreme results. Other views are crowded out. – People are reluctant to present their view if they believe that it is in the minority. – The majority will tend to disregard a minority view as being incorrect so that new information is ignored.  It may therefore be better to ask views individually  There is the “Eureka” situation where groups are good: when it takes several people to put together a solution (crossword puzzles) – the solution is seen immediately when it is suggested 4 Prediction Markets  Prices play the role of information in markets. You can get better results when people are sorted based on their own judgment of the value of their information – The Iowa experiment – Google  These are modeled after market economies and the “invisible hand.” 5 Overview 1. Markets as a metaphor for economics of organizational design 2. Centralization v. decentralization 3. Coordination 4. Decision making, hierarchy, & control 5. (next lecture) Job design & decision making 6. (next module) Incentives 6 1. Organizational Design of an Economy  Adam Smith » “… he intends only his own gain, & is … led by an invisible hand to promote an end which was no part of his intention … By pursuing his own interest he frequently promotes that of society more effectually than when he really intends to promote it.”  Leontief: central planning (centralization) is efficient – coordination, economies of scale, control  Hayek: market (decentralization) is more efficient – costly to move all info. to central planner; decentralization makes better use of specific knowledge of time & place: » “How can … fragments of knowledge existing in different minds bring about results which, if they were to be brought about deliberately, would require a knowledge on the part of the directing mind which no single person can possess?” 7 Markets as Information & Incentive Systems  Examples of markets as forms of organization – prediction markets (insurance, financial, etc.)  Market economies have 3 important features: – decentralization makes good use of “specific knowledge of time & place” – prices provide good “general knowledge” for coordination – incentives (through ownership) » » » » motivates good decision making moves decision rights to person with most valuable/ relevant specific knowledge motivates investments in human capital motivates creativity / innovation 8 Organizational Design of a Firm  Org. design must address the same problems  use of specific knowledge of time & place  coordination across decision makers  incentives for both  innovation & adaptation  Can we design an organization to mimic a market? – even if we can’t completely, the intuition is very useful  Note, though, the limits of markets – they are best at aggregating information (e.g., into prices or predictions) – when coordination in the sense of coordinated actions is important, organizations tend to be set up 9 2. Benefits of Centralization  Economies of scale – – – – physical capital managerial talent brand name & reputation design  Better use of central knowledge – aggregated information & experience of the combined organization  Better coordination – – – – – knowledge transfer across units consistency / standardization synchronization control common strategy 10 Benefits of Decentralization  Better use of specific knowledge dispersed throughout the organization  Prevents senior management from being overwhelmed  Training/ development & intrinsic motivation for lower level managers  Less bureaucratic/ more manageable scale 11 Specific Knowledge Cost of Transferring Knowledge Costly Specific Knowledge Cheap General Knowledge  Attributes of knowledge / information that make it more “specific” – costly to transfer » perishable » complex – costly to understand » requiring scientific or specialized technical skills » subjective or experiential – unreliable / risky to use » noisy (garbling) 12 3. Coordination & Structure  The classical approach: centralized hierarchy  Decentralization implies that coordination must happen at lower levels of the organization – roughly speaking, 2 kinds of coordination problems: “simple” & “integration”  Simple coordination: getting units to act in concert – real-time communication not needed, as long as actions of all units are compatible – use incentives, communication, job rotation, culture » e.g., UPS; 13 Modularization  Putting people with the most interdependent jobs together amounts to modularizing overall structure – ex: break XP Consulting into smaller divisions » regional? (NA; Europe; Asia) » type of customer? (Corporate; Government; Not-for-profit; etc.) » practice area? (Strategy; IT; 6s)  You can structure authority w/ some decisions organized by one set of divisions, others a different set » ex: decisions about hiring & compensation determined by region; decisions about training & promotion determined by practice area  Note how complex it starts to get … there are clear advantages to reducing overlapping lines of authority – hypothesis: the largest source of dis-economies of scale is bureaucracy (coordination costs) 14 Integration  Integration: for some decisions, pockets of specific knowledge throughout org. need to be combined – ex: Apple Computer laptop product design – use lateral mechanisms » teams, matrix » informal networks » e.g., product design CEO Engineering Sales Production Other B A – the most cumbersome organizational designs tend to involve integration problems A B B A  Or, balance the 2 goals of coordination & use of specific knowledge – separate decision management & control (below) 15 4. Decision Making, Hierarchy, & Control  Think of decision making as a 4-stage process 1. initiatives 2. ratification 3. implementation 4. monitoring Decision Management Decision Control (Hierarchy)  Different stages can be more centralized or decentralized 16 Notes on Decision Mgt. v. Control  It often makes sense to separate decision management from decision control – if decision maker has weak incentives » Board v. CEO – can provide benefits of decentralization & centralization at the same time » decentralizing decision management » centralizing decision control  The distinction is useful in practice – innovation process – managing change – empowerment 17 How Much Decision Control?  Consider 2 firms with 2 employees  The units evaluate new ideas differently – “Hierarchy”: W evaluates new ideas, passes some to G. G approves or rejects those – “Flat”: G&W both different new ideas Hierarchy Gladys Gladys Willie Flat Willie  N = # of ideas each can evaluate per period – flat firm evaluates twice as many ideas per period 18 Evaluating New Ideas  Assume new ideas are binary (good or bad / profitable or unprofitable)  At first stage, p = probability of correct decision; p > ½  At second stage (hierarchy only), q = probability of correct decision; q > p 19 Hierarchy Accepts p Good Accepts p∙q Rejects p(1-q) Rejects (1-p) Rejects (1-p) New Idea Accepts (1-p) Bad Rejects p Accepts (1-p)(1-q) Rejects (1-p)q Rejects p Willie Evaluates Gladys Evaluates Willie’s Recommendations 20 Flat Accepts p Good Rejects (1-p) New Idea Accepts (1-p) Bad Rejects p Gladys or Willie Evaluates 21 Results Flat Rate For One New Idea Accept Good Idea False Negative False Positive Reject Bad Idea Overall Throughput Accept Good Ideas False Negatives False Positives Reject Bad Ideas p 1-p 1-p p Hierarchy pּq 1-pּq (1-p)(1-q) 1-(1-p)(1-q) 2Nּp 2N(1-p) 2N(1-p) 2Nּp Nּpּq N(1-pּq) N(1-p)(1-q) N[1-(1-p)(1-q)] 22 Are Hierarchies Conservative? Most Rate For One New Idea Accept Good Idea False Negative False Positive Reject Bad Idea Overall Throughput Accept Good Ideas False Negatives False Positives Reject Bad Ideas Flat Hierarchy Flat Hierarchy > > > > Middle Hierarchy Flat Hierarchy Flat Flat Flat Flat Flat > > > > Hierarchy Hierarchy Hierarchy Hierarchy  Flat structures – – – – evaluate ideas more quickly evaluate more ideas for the same # of employees make more changes, good & bad have more successes & failures  What kind of environments favor a more hierarchical or flat structure? 23 Other Methods to Increase Control  Resources spent on accuracy (a & b)  Skills & emphasis of decision makers – liberal v. conservative evaluator – conservative org. likely to recruit / train more carefully  Incentives of decision makers – e.g., downside punishments & upside rewards  Constraints on decisions – e.g., budgets  Culture & process 24 Structure and Errors •Hierarchical •Reduce false positive and increase false negative •Approve fewer projects overall •Good where careful consideration is needed. Good with traditional industry; regulated industry. Bad for rapid change •Second Opinion - symmetrical upside and downside •Flat •Reduce fall negative and increase false positive •“Creative people not attracted to hierarchical firm” •Good when unprofitable projects are not too costly or when profitable projects are likely to be very profitable 25 5. Implementation  So what should XP Consulting consider in its structure?  First, Modularize overall structure, possibly in overlapping ways – ex: Cambridge Technology Partners – makes the problem more manageable – put most interdependent parts together, reducing coordination problems  Second, allocate decisions within each division: ask “who / what / where / when / why?” to identify key specific knowledge – – – – – who has valuable specific knowledge? what kind of knowledge? where in (& out) of the organization? when (is timing relevant)? why is it of economic value?  Third, think about what needs to be made consistent or coordinated across the division or whole organization 26 Implementation  The last two give strong guidance on what to decentralize & centralize  Fourth, go back & refine the overall structure – try to streamline further to cut bureaucracy – look for & address coordination problems – integration problems require the most attention, & will create most of your day-to-day headaches  Fifth, design jobs (next lecture) – balance benefits of specialization, standardization against benefits of using specific knowledge, intrinsic motivation  In all of this, balance desires for control v. creativity & adaptation – self organizing systems can be extraordinarily powerful; don’t be a control freak!  Sixth, design performance evaluation & incentives to match job design (after Midterm) 27 6. Economic Ideas  The “knowledge problem” of organizational design – specific knowledge – coordination types & mechanisms – incentives & price mechanisms  Decision making – decentralization v. centralization – decision management v. control – degrees of decision control / hierarchy, & their effects 28 Summary Points  The metaphor of a market highlights the role of economics in organizational design – design is largely about creating & making use of knowledge » by its nature, specific knowledge tends to have more economic value – incentives play a crucial role » approximating ownership » performance measures are “prices” – but market approaches are limited when complex coordination (especially of the “integration” kind) is needed  The concepts apply to design of an individual job, to a workgroup, to structure of a global conglomerate 29