TAC SCM Survey

1
TAC SCM Survey

• Alberto Sardinha
• Norman Sadeh
• e-Supply Chain Management Lab
• Carnegie Mellon University

2
Outline

• Results of the TAC SCM Survey
• 55 Respondents from
• From 30 different organizations
• From 15 different countries
• The survey was conducted between May 20, 2006 and
  June 3, 2006.

3
Question 1

• Enter your name and that of your organization

4
Number of Respondents per Organization
5
Question 2

• Indicate each of the years when you had an entry
  in the TAC-SCM tournament

6
Indicate each of the years when you had an entry
in the TAC-SCM tournament
7
Question 3

• If you had entries in 2005 or earlier but didnt
  enter the competition in 2006, what was (were)
  the main reason(s)?

8
If you had entries in 2005 or earlier but didnt
enter the competition in 2006, what was (were)
the main reason(s)
9
If you had entries in 2005 or earlier but didnt
enter the competition in 2006, what was (were)
the main reason(s)

• Most common answers
• Busy Schedule
• Joined another team
• The current game does not provide enough
  opportunities for machine learning

() This is a subset of the answers. The full
text is provided separately
10
Question 4

• Assuming that the 2007 competition takes place in
  July-August, which of the following would
  increase/decrease the chances you participate?

11
Which of the following would increase/decrease
the chances you participate?
12
Which of the following would increase/decrease
the chances you participate?
13
Which of the following would increase/decrease
the chances you participate?
14
Which of the following would increase/decrease
the chances you participate?
15
Which of the following would increase/decrease
the chances you participate?
e.g. semi-finals with 24 agents and finals with
12 agents, being mixed and matched for a couple
of days
16
Which of the following would increase/decrease
the chances you participate?
e.g. 15-20 minute games instead of 55 minute
games - with possible provisions to reduce
startup/shutdown effects
17
Which of the following would increase/decrease
the chances you participate
18
Which of the following would increase/decrease
the chances you participate?
e.g. combination of short-term and long-term
procurement contracts
19
Question 5

• Please use this box to provide additional details
  on any of your answers under Question 4

20
Sample of Responses - I

• Improve agentware documentation
• "Start-Kit" for new competitors
• Improved API if possible to assist agent
  modularity
• Donations and symbolic entry fees instead of 250
• More complicated factory problem
• e.g. switch-over costs, multiple resources,
  machine breakdown perhaps correlated with high
  utilization
• Double the game length to make it more realistic
  and to reduce start and end game effects
• A more complex reputation management

21
Sample of Responses - II

• Introduction of multiple leagues
• Experimental" league with a new set of rules
• League with 2006 rules
• Undergraduate league
• League with "simpler" specifications
• The current supplier model encourages early
  component orders and discourages the use of
  negotiation strategies

() This summary provides a small
portion/highlights of all the comments. The
full text is provided in another file.
22
Sample of Responses - III

• Availability of game data should be delayed till
  after the tournament is over
• More opportunities for learning
• Competition should be designed to test security
  aspects of agents
• Finals with 12 agents
• Reduce predictability on the customer side
• Introduce a loan limit
• Products could have zero demand for several days
  at the beginning or end of games e.g. product
  launch and phase-out
• Allow manufacturers to respond to consumer bids
  with offers for substitutable products (e.g.
  different CPUs, etc)

23
Question 6

• Please briefly describe your main research
  interests and motivations for participating in
  TAC-SCM
• e.g. adaptive techniques, optimization,
  Supply Chain Management, game theoretic
  issues, etc.

24
Please briefly describe your main research
interests and motivations for participating in
TAC-SCM
() First topic listed even though respondents
were not instructed to list topics in any
particular order
25
(No Transcript)
26
Question 7

• Do you use TAC-SCM software in courses that you
  teach?

27
Do you use TAC-SCM software in courses that you
teach?
28
Do you use TAC-SCM software in courses that you
teach?
29
Do you use TAC-SCM software in courses that you
teach?
30
Do you use TAC-SCM software in courses that you
teach?
31
Do you use TAC-SCM software in courses that you
teach?
32
Question 8

• Do you find the TAC-SCM Agent repository useful?

33
Do you find the TAC-SCM Agent repository useful?
34
Question 9

• Do you have any suggestions for improving the
  research/educational value of the competition?

35
Samples Responses - I

• The competition doesnt encourage opponent
  modeling
• The game does not provide detailed information
  about the behavior of individual agents.
• Dont release game logs but make more information
  reports available to agents to somewhat
  compensate
• Moving TAC-SCM 2007 to July would increase
  participation
• We already make use of available data logs from
  previous games. This has to continue.
• The rules of the game are too complex for
  students taking an agent course to try to make
  their own agents.
• Perhaps, we could develop a lighter version that
  will pick the students' interest enough to entice
  some of them to develop a full version agent.
• Changes to the game rules will be needed to keep
  the research papers coming

() sample responses. The full text is provided
in another file.
36
Samples Responses - II

• A more sophisticated game viewer will be very
  interesting for educational purposes.
• I think having some support from industry is very
  important. Especially in credibility.
• Making all the agents available after the
  competition
• It's difficult to justify dedicating significant
  resources to the SCM game when you work for an
  organization that needs real impact and results.
• Increase the sharing of agent binaries, source
  code, analysis tools, agent descriptions etc.
• The ideal specification should have both high
  variability and moderate predictability.
• Long-term "strategic" decisions
• Games could last for 20 quarters (5 years), with
  each quarter taking 1 or 2 real minutes.
• A minor point about the current specification is
  that supplier prices have a rather strange
  pattern, in that it is often possible to get
  components for cheaper in the very short term
  (2-3 days out) than in the mid term (6-10 days
  out).

() This summary provides a small
portion/highlights of all the comments. The
full text is provided in another file.
37
Samples Responses - III

• Developing a GUI front-end to develop agents.
  This is critical for many Industrial Engineering
  students and Business School students that lack
  programming skills.
• The possibility of market shocks (i.e. suppliers
  going out of business or customer demand
  collapsing in one market segment)

() This summary provides a small
portion/highlights of all the comments. The
full text is provided in another file.
38
Question 10

• Enter here any other comments you may have about
  TAC-SCM and how you would like the tournament to
  evolve over the years to come

39
Samples Responses - I

• I am imagining right now is an environment which
  houses lots of agents
• Having TAC run through late April was very
  unfortunate for my team
• Reduce the registration fee for the teams who
  lost in the previous year. It is quite difficult
  for us to get the funding for next year if we did
  not do well in the finals.
• Add a secondary market for components and
  finished goods. Allow agents to buy other agents.
  
• Final rounds should include more games, more
  representative
• Attracting more participants could be done in 2
  ways
• advertising this contest in the academic
  environment
• offering some cash prizes for the winners even if
  symbolic
• The requirement to have exactly 6 agents in a
  game is a bit of a problem for tournament
  organization.
• Perhaps the number could be variable? Or perhaps
  the technique used this year of having a "filler"
  agent is a reasonable compromise.
• Introduce mix of short-term and long-term
  contracts
• Introduce a one-day lag (or greater) before
  detailed game logs are accessible to competitors

()The full text is provided in another file.