the bullwhip explorer

my professional homepage

google

 

Intro Bibliography Book Shop The Bullwhip Explorer Management Games Excel Add-in The Super-ellipse Explorer Links Bullwhip Help / Contact Info

 

This page now has 3 different java based bullwhip explorers by Takamichi Hosoda.   The first explorer models a three echelon supply chain and investigates altruistic behavior or local verses global optimization.  The second explorer, models a two echelon supply chain that uses minimum mean squared error forecasting.   The final explorer highlights the impact of different forecasting methods in a single supply chain echelon.  There is also a list of other bullwhip simulators on the internet at the end of the page.  If you would like to contribute an explorer, please email me.  

 

A Simulation of a Three Echelon Supply Chain with Altruistic Behavior by Takamichi Hosoda.

How to enjoy this model:

STEP 1. Define the demand pattern by selecting an autoregressive parameter then click the "sim" button at the bottom to see the results.

STEP 2. Observe the time-series of the order and net stock levels and the "SD (Standard Deviation)" at the top of each graph. Those SDs are obtained from the time-series data you see in the graphs. Every time you click the button, the new demand pattern will be generated and you can see different results even if the same parameter setting you use. Please crick many times and see what happens!

STEP 3. Go back to STEP1 and try different value of the parameter.


Sorry, you do not have Java support


Model Description:

Demand pattern: First-order autoregressive pattern (AR(1)) with the error term following an N(0, 1) distribution is assumed. Even though the autoregressive parameter can be negative in this model, it is common to have positively auto-correlated demand pattern. Try some values and see the power of expression of AR(1) process!

Traditional Supply Chain (Gray line): Each player uses the same ordering policy, which is the order-up-to policy with the MMSE (Minimum Mean Square Error) forecast method. This supply chain is, therefore, a sequence of the optimum policies. This model has been studied before in Hosoda and Disney (2006).

Generalised Supply Chain (Black line): The first TWO echelon players, the retailer and the distributor, employ the order-up-to policy with a proportional controller added into the inventory position feedback loop. The third echelon player, the manufacturer only uses the traditional order-up-to policy. All players adopt the MMSE forecasting method. From this model, by comparing to the traditional SC, you can see the altruistic behavior at  the retailer brings the benefit for overall supply chain: lowering the bullwhip effect at every stage, and lowering the net stock variances at the second and third echelon players. In terms of net stock, Hosoda (2005) has shown that in infinite horizon, the benefit from the upper two echelon is large enough to compensate for the loss at the first echelon, and also has shown that the bullwhip is mitigated. An important insight from this model is that a significant amount of benefit comes from each player doing what is the best for itself and the supply chain, rather than doing what is the best for its own interests. In other words, a sequence of the optimum policies will not bring globally optimal values to the overall supply chain anymore.

 

Notes:

Acknowledgement:

This model is created by Takamichi Hosoda, Logistics Systems Dynamics Group, Cardiff Business School , in June 2005, as a part of his PhD research.

References:

 

A Simulation of a Two Echelon Supply Chain by Takamichi Hosoda

How to enjoy this model:

STEP 1. Define the demand pattern by selecting an autoregressive parameter using the scroll bar on the left.

STEP 2. Set the replenishment lead-times at each echelon.

STEP 3. Click the "sim" button at the bottom to run the model.

STEP 4. Observe the time-series of the order rate and inventory levels and the "Variance Ratio" at the top of each graph. "Variance Ratio" is the ratio of the demand variance and the order/net stock variance. If "Variance Ratio" is greater than 1, then bullwhip has occurred. Every time you click the button, new time-series data is generated and you can see different results. Please click many times and see what happens!

STEP 5. Go back to STEP1 and try new value settings.


Sorry, you do not have Java support


Model Description:

Demand pattern: First-order autoregressive pattern (AR(1)) with the error term following a N(0, 1) distribution is assumed. Even though the autoregressive parameter can be negative in this model, it is common to have positively auto-correlated demand patterns (Lee et al., 2000). Try some values and see the power of expression of AR(1) process!

Ordering policy: Each player uses the order-up-to policy with the minimum mean square error forecasting scheme.

Notes:

Every time you click the button, the ranges of Y-axis are automatically adjusted in the same scale so that you can compare the magnitude of variations visually.

The average values of all time-series are converted into zero.

Acknowledgements:

This model is created by Takamichi Hosoda, Logistics Systems Dynamics Group, Cardiff Business School , in July 2004, as a part of his PhD research. If you are interested in the mathematical backgrounds of this model, please take a look at Hosoda and Disney (2004). This paper also provides the equations of variance ratios in the infinite time-horizon and some practically useful insights in supply chain management. Any feedbacks are welcome.

References:

 

A Simulation of a One Echelon Supply Chain with Different Forecasting Methods  by Takamichi Hosoda

How to enjoy this model:

STEP 1. Define the demand pattern by selecting an autoregressive parameter using the scroll bar.

STEP 2. Finalize the forecast models by selecting an EWMA (Exponentially Weighted Moving Average) parameter and an MA (Moving Average) parameter using the scroll bars.

STEP 3. Set the replenishment lead-time.

STEP 4. Click the "sim" button at the bottom to run the model.

STEP 5. Observe the time-series of the order rate and inventory levels and the "Variance Ratio" at the top of each graph. "Variance Ratio" is the ratio of the demand variance and the order/net stock variance. If "Variance Ratio" is greater than 1, the bullwhip has occurred. This model will highlight the impact of the forecasting method on the "Variance Ratio". Every time you click the button, the new time-series data are generated and you can see different results. Please click many times and see what happens!

STEP 6. Go back to STEP 1 and try new value settings.



Model Description:

Demand pattern: First-order autoregressive pattern (AR(1)) with the error term following an N(0, 1) distribution is assumed. Even though the autoregressive parameter can be negative in this model, it is common to have positively auto-correlated demand pattern (Lee et al., 2000). Try some values and see the power of expression of AR(1) process!

Ordering policy: Each player uses the same ordering policy, the order-up-to policy but with different forecasting methods:
1) MMSE (Minimum Mean Square Error) forecast, 2) EWMA (Exponentially Weighted Moving Average) forecast, and 3) MA (Moving Average) forecast.

Notes:

Acknowledgement:

This model is created by Takamichi Hosoda, Logistics Systems Dynamics Group, Cardiff Business School , in July 2004, as a part of his PhD research. If you are interested in the mathematical backgrounds of this model, please take a look at Hosoda and Disney (2004a), and Hosoda and Disney (2004b). Those paper also provide the equations of variance ratios in the infinite time-horizon and some practically useful insights in supply chain management. Any feedbacks are welcome.

References:

 

Other bullwhip games and simulations on the web