Beer game

"Carranza, Octavio" · Post by **"Carranza, Octavio"** » Tue Apr 02, 1996 4:27 pm

Jim Hines · Post by **Jim Hines** » Wed Apr 03, 1996 10:02 am

Octavio,

The step increase is the classic pattern and is desirable because it
shows so clearly that the oscillations do not come from the input. If
you change the input pattern you would want to maintain this feature
(Are there any other test inputs that have this character? Maybe a ramp
or a step down. A pulse or a step up-and-down have in them the pattern
that many people will see (an increase and a decrease) I am not sure
that a step-down will elicit the response as effectively).

Even more to the point. We often have people who have played the game
before in our sessions (which tend to be pretty big). We simply set up
an "experts" table. They expect a different input. We give them the
same one (or actually one that has a bigger step) -- the same results
ensue. This allows the additional point to be made that it is very
difficult to learn how to control dynamic systems.

Regards,
Jim Hines
LeapTec and M.I.T.
jimhines@interserv.com

Bill Harris · Post by **Bill Harris** » Thu Apr 04, 1996 2:54 pm

Jim Hines wrote:

>(Are there any other test inputs that have this character?

While I havent tried it in a classroom presentation, you could also pick a
constant (i.e., no step) input at a level that doesnt match the initial
conditions in all of the intermediate levels (that is, have a constant
order level of 8 throughout the simulation). That should prang the
system, too, and should generate similar dynamics. (Admittedly, I havent
tried this one, but Ive tried a computer simulation of a similar system
with similiar initial conditions, and it worked.) The disadvantage for
inexperienced users would be that the dynamics start early, so it is harder
to run the 4 weeks of practice effectively. It might work quite well for
experienced users, as the game board looks the same, and the inputs look
deceptively easy. You certainly couldnt blame the systems dynamics on
the customer inputs then!

Id try that in a game before putting it in a course.

Bill

--
Bill Harris Hewlett-Packard Co.
R&D Productivity Department Lake Stevens Division
domain: billh@lsid.hp.com M/S 330
phone: (206) 335-2200 8600 Soper Hill Road
fax: (206) 335-2828 Everett, WA 98205-1298

saeed@ait.ac.th (Professor Khali · Mon Apr 08, 1996 2:24 pm

I have used two sets of orders for groups in the same class, a step (4 to

and random noise (with mean 4 and std dev 1). The behavior is similar in
the two cases, and it generates a good discussion about internal trends vs
external disturbance.

Khalid

Khalid Saeed
Professor and Program Coordinator
Infrastructure Planning & Management Program
School of Civil Engineering
ASIAN INSTITUTE OF TECHNOLOGY
G.P.O. Box 2754, Bangkok, THAILAND
phones: (66-2)524-5681, (66-2)524-5785; fax: (66-2)524-5776
email saeed@ait.ac.th

Visit our program website at: http://www.ipm.ait.ac.th/

jsterman@MIT.EDU (John Sterman) · Sun Oct 06, 1996 7:54 pm

Octavio Carranza want to know what demand pattern can be used with
people who have already played the beer game once.

We have had excellent experiences using an order deck with the pattern
4,4,4,4,8,12,12,12.....12

This is a step input just as in the base case, but the step height is
larger. Most people wo have played the game before dont learn a robust
ordering rule, but instead remember that demand stayed constant at 8 and
that they shouldnt over-react. Often they immediately jump their
orders to 8 throughout the supply chain without even waiting for an
increase from their customer. When they find their forecast of 8 proves
to be wrong, they try to hold the line on orders, resulting in a huge
backlog. Eventually, they crack under the strain and then order a lot.
Usually one observes that by the end of the game these folks are
generating a cycle that is largely indistinguishable from the cycle
generated by first time players. The large backlogs lead to high costs,
so these experienced teams almost never win. The experience is a good
lesson in the need to focus on underlying structure and the development
of improved decision rules for managing a system rather than focusing on
trying to improve your forecasting - inevitably, the forecast will be
wrong, and what then?

John Sterman
jsterman@MIT.EDU