FOLLOWUP Beer/Wine Game

[Niall Palfreyman]

Hi,

I finally ran the computer-simulated Beer Game with a group of students
yesterday, and came out with a whole load of surprises and learnings.
Since several people on the list asked me to report on the event, I
thought Id also take advantage of the opportunity to ask for advice on
a couple of issues. I should mention that I have never before personally
run or participated in the Beer Game, which may explain why some of my
expectations turned out to be inappropriate. First a few words on how I
ran the game.

I originally wanted to get away from the alcoholic content of the game,
and called it the Mineral Water Game. Unfortunately, both mineral water
and beer are here such everyday commodities that the long delivery times
seemed a little unbelievable. As a consequence I in the end called it
the Wine Game, and adjusted the parameters of the game slightly: orders
took 1 week instead of 2 weeks to arrive at the supplier, and delivery
took 2 weeks - total of 3 weeks turnaround time from delivery to order.
As I understand it, the turnaround time in the Beer Game is a total of 4
weeks. To still keep the possibility of interesting dynamics, and also
to incorporate a larger group of students, I extended the supply chain
to 5 (instead of 4) links: Manufacturer, Bottler, Distributor,
Warehouse, Supermarket. I tested these parameters with the Stella model
enclosed with this mail [write directly to the author for this
mailto:niall.palfreyman@fh-weihenstephan.de - not the list],
and got a satisfying set of wild cycles
reverberating from the single customer increase from 2 to 4 crates in
the 5th week.

I ran the game with a class of 30 students split into 6 groups of 5
people. 5 of these groups played the roles of the various links in the
supply chain; 1 group played the role of the mediators between the
groups. In each week of the simulation, the mediators brought me this
weeks order from the respective links, and I entered these orders into
the Stella model. I then clicked on the Run button to step the
simulation to the next week, and the mediators then took the next weeks
results out to the separate links. Each link received information each
week on: last weeks delivery of wine, unfulfilled orders, customers
new order, current inventory, and costs so far. The Stella simulation
also kept track of a number of interesting variables for the discussion
at the end.

My intention in using the simulation was to accelerate the game, and in
this I was successful. The preplay briefing lasted about 20 minutes, the
game run lasted about 70 minutes, and the final debriefing discussion
took about 30 minutes. The mood in the room was pleasant and talkative,
but without the shouting I had expected from other reports of the game.
The students enjoyed the game, and reported positively on the learning
from the game. At the same time there were certain problems which Ill
report below.

I have included a zipped version of the model I used to run the game
with this mail. Please feel free to use it, and Id be glad to get any
feedback from people on the use of it. You can run the game solely from
the controls and graphs on the first (opening) screen. The simulation is
ready to run: the pause interval is already set at 1 week for stepping
from one week to the next, and Nialls thirst (on the second screen
scrolling downwards) is set to the standard step function from 2 to 4
crates in the 5th week. The 5 slider controls for setting the current
orders are currently set to manual mode for entering the players
orders, but can be set to formula mode to eliminate one or more links
from the supply chain.

PROBLEMS:
There was one problem with the game run which almost destroyed the
learning effect. Im still not sure where this problem came from, and
Id be interested to hear opinions from more experienced people.
Basically, the problem was that we never got the avalanches of
late-arriving deliveries which I had come to expect from other reports
on Beer Game runs. We never hit the wild cycles which are supposed to
frustrate all the players so much. We got some (fortunately sufficient
to feed into the final debriefing), but not nearly as much as I had
expected. I can think of four possible contributing factors for this:

1. Right from the beginning, there was no smooth flow of
2-crates-per-week deliveries which I had expected. I had expected smooth
flow at the beginning, which is then broken up from week 5 by the first
4-crate delivery, but there was wild ordering from the word "go". Im
wondering if I missed out something important from the pregame briefing
which would have encouraged them to start off more smoothly?

2. The manufacturers, right from the start, built up an inventory of
about 30 crates, foreseeing that they might need to buffer their orders.
Is this usual? The result was that they effectively eliminated
themselves as a source of fluctuations in the system, since they always
delivered reliably. Should I have avoided this by some preliminary
comment?

3. There may well be a fault in my Stella model. We never really built
up any great backlogs of orders, and a couple of teams reckoned that
some orders never got delivered. Id be very grateful if any of you
could find such a mistake, since I do believe that the principle of
running the game via a simulation is a good one.

4. Finally, one mistake I made was to forget to tell the mediating group
to keep quiet about the number of weeks of the simulation. All groups
quickly found out that the simulation only ran for 36 weeks, and so ran
down their inventories towards the end of that time.

All in all, the experiment was successful, and I very much want to do it
again, but I would like to solve these problems beforehand, so any tips
would be very welcome!

Best wishes,
Niall Palfreyman.
From: Niall Palfreyman <niall.palfreyman@fh-weihenstephan.de>