Decision Making by juanagui

VIEWS: 6,351 PAGES: 29

									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                                                                            Cheap

Specific Knowledge                                                    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
                                                                            CEO
   – ex: Apple Computer laptop product design
   – use lateral mechanisms
                                                      Engineering   Sales         Production   Other
      » teams, matrix
      » informal networks
                                                             B          A
      » e.g., product design
   – the most cumbersome organizational designs              A                           B

     tend to involve integration problems                               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

                                   Decision Management
     1. initiatives
     2. ratification
     3. implementation
     4. monitoring                   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                       Hierarchy            Flat

  The units evaluate new ideas differently
                                                             Gladys
    – “Hierarchy”: W evaluates new ideas, passes
      some to G. G approves or rejects those
                                                                       Gladys          Willie
    – “Flat”: G&W both different new ideas
                                                             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∙q
                               Accepts
                                  p
                                                       Rejects
                  Good                                 p(1-q)

                                Rejects                Rejects
                                 (1-p)                  (1-p)


       New Idea
                                                       Accepts
                               Accepts                 (1-p)(1-q)
                                (1-p)
                                                       Rejects
                                                       (1-p)q
                  Bad

                                Rejects                Rejects
                                   p                      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       Hierarchy
     Rate For One New Idea
         Accept Good Idea    p         pּq
         False Negative      1-p       1-pּq
         False Positive      1-p       (1-p)(1-q)
         Reject Bad Idea     p         1-(1-p)(1-q)

     Overall Throughput
         Accept Good Ideas   2Nּp      Nּpּq
         False Negatives     2N(1-p)   N(1-pּq)
         False Positives     2N(1-p)   N(1-p)(1-q)
         Reject Bad Ideas    2Nּp      N[1-(1-p)(1-q)]



                                                         22
Are Hierarchies Conservative?
                                                          Most           Middle
                    Rate For One New Idea
                        Accept Good Idea                   Flat     >   Hierarchy
                        False Negative                  Hierarchy   >      Flat
                        False Positive                     Flat     >   Hierarchy
                        Reject Bad Idea                 Hierarchy   >      Flat

                    Overall Throughput
                        Accept Good Ideas                 Flat      >   Hierarchy
                        False Negatives                   Flat      >   Hierarchy
                        False Positives                   Flat      >   Hierarchy
                        Reject Bad Ideas                  Flat      >   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

								
To top