Behavioral Causes of
the “Bullwhip” Effect in
cor Supply Chains
Rachel Croson <a:e
eeeee8
Karen Donohue | eeee
eee0e@
Elena Katok | eee
ee
J ohn Sterman
The “Bullwhip” Effect 332°
e Orders to increase in variation as one moves
up a Supply chain.
e The effect is costly because it causes
excessive inventories, poor customer service,
and unnecessary capital investment.
Operational Causes oe:
e There is a great deal of research on
operational causes of the bullwhip effect (see
for example Lee et al. 1997):
e demand signal processing,
e inventory rationing,
e order batching
e price variations
Behavioral Causes of the $325,
Bullwhip Effect ane
“... the key to improved performance lies within
the policy individuals use to manage the
system and notin the external environment.
Even a perfect forecast will not prevent a
manager who ignores the supply line from
over ordering.” (Sterman 1989, p. 336).
e Implication: the Bullwhip effect will persist
even if ALL operational causes are removed
(even with constant and known demand).
The “Beer Distribution Game”
e A vehicle we use to study the bullwhip effect in the
laboratory.
Your Role is: Retailer
This is the beginning of week: 1
CUSTOMER PRODUCTION
* os os: os of
al ee _ a _|
12 SHIPMENTS SHIPMENTS SHIPMENTS PRODUCTION
ve ..... nn Lt | ia ne
| sed — a =
Retailer Wholesaler Distributor Manufacturer
Research Questions sss
e Will the bullwhip effect persist in an
environment with constant and known
demand?
e If so, then we can separate possible causes
into two broad categories
e Cognitive limitations
e Inability to coordinate
Experimental Design oe:
e Compares performance of subjects in the
Same roles in teams with all human
participants, to teams with one human
participant.
e If we see improved performance in the
automated teams, we can conclude that, at
least partially, the problem is due to the
inability to coordinate.
Experimental Design
Team Composition
Know Optimal All Human One Human per
Policy Teams Team
YES |=12,5teams |l =12, 20 teams
|=0, 5 teams | =0, 20 teams
NO | =12, 5 teams
1=0, 5 teams
Customer demand is constant at 4; this is public information
There are 4 cases in each delay position
| = Initial Inventory is either 0 or 12, depending on the treatment.
One Example
Cases
500
1 3.5 7 9 11131517 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47
Period
—t— Retailer —#— Wholesaler —&— Distributor —®— Manufacturer
All human team
* On-hand inventory
* Initial inventory =0
* No information about
optimal policy
provided
Comparisons by Role
Retailers
Distributors
Automated
ees Human
Wholesalers
Manufacturers
Overall Performance...
1000000
100000 +
10000 +
1000 +
100 +
10 +
Average Cost per Team
14
Initial Inventory
m Automated a@lnformation ONo Information
Estimating Behavior
From Sterman '89:
Order =max {0, EO +af{1* 4) -b(SI
Where:
EO =expected order
I* = target inventory
SL * =target supply line
| = actual inventory
SL =actual supply line
a and b are adjustment parameters to be estimated
Ignoring supply line...
Average b
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
12% 68%
51% 52%
° 45% :
31%
Initial Inventory
mAutomated ailnformation oO No Information
Conclusions ese
e The bullwhip effect persists with known and
constant demand.
e Behavioral explanation
e Telling subjects what the optimal ordering
policy is does not help them.
e Human subjects do better when other team
members are computerized than when the
other team members are human.
e Coordination Is part of the story
Back to the Top
