demand uncertainty

The Supply Chain Uncertainty Circle

Daniel Dumke — Wed, 27 Oct 2010 16:17:00 +0000

How to shrink the “Uncertainty Circle” is the topic of a paper I read today. It has been written by Rachel Mason-Jones and Denis R. Towill and can be downloaded here free of charge.

This article from 1998 focusses on how to shrink the uncertainty within a supply chain and therefore improve performance, since

we believe that those companies which cope best with

uncertainty, are most likely to produce internationally

competitive bottom-line performance.

This hypothesis is also backed by several other authors like, Craighead et al. (2009) or Hendricks and Singhal (2005).

Within the product delivery process for a single company, the authors find four causes of uncertainty: supply side, manufacturing processes, control systems and demand side (see figure 1).

Causes of Supply Chain Uncertainty (Mason-Jones and Towill 1998)

The Circle

With the Uncertainty Circle it is now possible to visualize to effect of shrinking one of the risks without the other: Overall risk is only reduced by a (comparatively) small amount. This furthermore implies the goal for supply chain risk managers: Reduction of risk on all frontiers!

Uncertainty Circle and Risk Reduction Process (Mason-Jones and Towill 1998)

This of course does not have to happen at the same time. Companies usually start with improvements in their own manufacturing processes and establish lean thinking to improve lead times and quality. As a next step this concept may be enlarged to the companies suppliers as well to reduce supply side risks.

The authors then argue that much potential is left untouched on the reduction of demand side uncertainties and control systems.

Demand side risks can be reduced by improving the information flow from the end-customer throughout the supply chain and companies can tackle control risks by improving their ability to act on this new information with a adequate Decision Support System.

Supply Chain Design Strategies

The authors then simulate a given supply chain and test strategies to improve (in figure 3) “Information pipeline time compression only”, “Material flow pipeline compression only”, “Total Cycle Time compression”.

Rating Risk Mitigation Strategies (click to enlarge; Mason-Jones and Towill 1998)

Conclusion

Even though the article has been published twelve years ago the concept of the Uncertainty Circle can still be used as a framework to reduce risks within a supply chain. The comparison table shows the effects of different risk mitigation strategies clearly and should adapted and used by more companies.

Reference:

Mason-Jones, Rachel, & Towill, Denis R. (1998). Shrinking the Supply Chain Uncertainty Circle IOM Control, 24 (7), 17-22

Craighead, C., Hult, G., & Ketchen Jr., D. (2009). The effects of innovation–cost strategy, knowledge, and action in the supply chain on firm performance Journal of Operations Management, 27 (5), 405-421 DOI: 10.1016/j.jom.2009.01.002

Hendricks, K., & Singhal, V. (2005). An empirical analysis of the effect of supply chain disruptions on long-run stock price performance and equity risk of the firm Production and Operations Managment, 14 (1), 35-52

Rate This:

Tags:

bullwhip effect

demand uncertainty

lean

mitigation strategies

network design

risk

framework

Simulation of Supply Chain Disruptions

Daniel Dumke — Wed, 22 Sep 2010 13:06:00 +0000

Still too many cooperations do not analyze their supply networks using consistent and scientifically proven methods. Some already do. One case of a company (ABC) is described below.

Goals and Methods

ABC company wanted to know more about their exposure to supply chain disruptions originating from their own plants but also the connected transportation links, suppliers and customers. Specifically, the goals were:

Assess the current level of supply chain disruption risk in the system
Test different mitigation strategies
Provide a tool that can be used in the case of a disruption to validate recovery steps before putting them into action
Identify redundancy in the system that can be removed without affecting the risk level

The authors therefore designed a supply chain model for two major products (one high, one low volume product) using discrete event simulation in Arena, one of the most often used simulation environments. To obtain distributions for the resulting performance measures the authors take advantage of Monte Carlo simulation using an excel tool called @Risk.

Simulation

The demand and risk patterns were evaluated using available data and expert interviews. In the second step mathematical distribution functions were adjusted to fit the empirical observations / forecast to input those into the model.

The company was very concerned with the effects of the disruptions on their customer service, so the Fill Rate was chosen as the relevant performance measure.

Results

The model was analyzed in two ways: First in an steady state analysis the current risk levels in the model were analyzed. In a second step the model was adjusted, by implementing one of the following strategies: implementation of inventory control and improvement of sourcing.

Even though only one simple measure was used, the ABC company was able to gain valuable insights into their supply chain.

Status Quo: All runs of the risk model experienced at least one week where 0% of the incoming orders were satisfied for the available inventory levels (two to six weeks)
Mitigation Strategies: The output clearly showed that reducing response times or increasing capacity of back-up methods were able to reduce the impact of a disruption and speed recovery.
ABC realized that they lack a database with historic distribution information for disruptions
There were no formal strategic-level mitigation procedures
Risk exposure is very dependent on the current state of the models parameters (eg. if a disruption occurs in a low inventory state, recovery lasts much longer)

The most important result is that the process of implementing such a simulation model can help the company think about supply chain disruptions more intensively and such a project can be used as a starting point for systematically documenting the risks and contingency plans in case of a disruption.

Reference:

Schmitt, Amanda J., & Singh Mahender (2009). Quantifying Supply Chain Disruption Risk Using Monte Carlo and Discrete-Event Simulation Proceedings of the 2009 Winter Simulation Conference, 1237-1248

Rate This:

Tags:

Disruption-Management Strategies for Short Life-Cycle Products

Daniel Dumke — Sat, 10 Jul 2010 07:05:00 +0000

In his 2009 paper Brian Tomlin analyzes strategies to mitigate disruption risks in a three echelon supply chain.

Setting

Focus in his research is a single company, with its suppliers and customers. The objective is to maximize expected utility, while demand and supply are uncertain. There are two products available which can be used as substitutes. The time horizon for the decision maker is one season where the products can be sold.

Three different sourcing structures are considered.

Different sourcing structures (Source: Tomlin 2009)

Variables / Attributes of the SC

The following attributes in the supply chain are used as free variables and can therefore be modified:

Suppliers
- Supplier reliability
- Correlation between supplier failures
- Proportion of order cost charged in the event of a failure
- Relative cost of emergency supplier
Products
- Contribution margin
- Demand uncertainty
- Demand correlation
- Substitutability
Firm
- Risk aversion of the decision maker

Strategies

The effects of three major adaption strategies on the risk exposure / expected profit are tested:

Supplier diversification
adding another supplier to lower the overall risk for failure
Contingent sourcing
adding a supplier which is only used in case the main supplier fails to deliver
Demand switching
shifting customer demand to a different product

There is also the possibility not to engage in any activity and accept the risks.

Results

It is shown that contingent sourcing is the preferred strategy to supplier diversification for the firm, if the failure probability of the supplier increases. Furthermore, demand switching is an effective strategy for controlling demand risk and therefore is the preferred strategy with low supply risk.

Higher levels of risk aversion lead to a preference for contingent sourcing over the other strategies.

Conclusion

Tomlin shows viable and effective strategies to increase profit (utility) in supply chain settings with disruption risk. He therefore calculates the expected profits in different situations and uses the results to give suggestion for situational optimal strategies to mitigate these risks.

Reference:

Tomlin, B. (2009). Disruption-management strategies for short life-cycle products Naval Research Logistics, 56 (4), 318-347 DOI: 10.1002/nav.20344

Rate This:

Tags:

demand uncertainty

mitigation strategies

Design of Agile Supply Chains

Daniel Dumke — Tue, 29 Jun 2010 13:28:00 +0000

I already reviewed two other articles about agile supply chains. One on the role of distribution centers in supply chains and one on the migration from lean to agile supply chains.

But the question comes to mind if lean and agile supply chains have to be mutually exclusive or if it is possible to combine them. Christopher and Towill (2001) had a look at this question regarding the supply chain design.

Factors for Leanness / Agility

Several factors influence the design decision if a supply chain should be more agile or lean. Nowadays more often than not supply chains compete, not companies. This implies that not only companies must act market orient, but also supply chains.
Market orientation can be analyzed using the market winner / market qualifier concept. This also means that there are different foci of the lean and agile supply chain.

Correlation of Agile / Lean Supply Chains to Market Winners / Market Qualifiers

Strategies

Christopher and Towill mention three ways to combine the lean and agile paradigm in one supply chain:

80/20 approach
Usually in a supply chain with multiple products only a few of them make most of the overall volume (20% make 80% of the volume). Result: for the high volume products choose a lean approach, while for the low volume products an agile strategy might be better suited.
Decoupling point approach
Strategic inventory can act as a decoupling point within a supply chain. If you select this point wisely, it can be beneficial to use a lean strategy up to this point and an agile strategy after this point.
Base and surge demand
the strategies can also be combined if the demand pattern shows a base demand. This demand could be produced using a lean strategy, while the surge implements a more agile approach.
Base and Surge demand
Source: Christopher and Towill (2001)

Conclusion

It is important to see, that not only leanness and agility can be combined, but also the strategies mentioned above can be used complementarily.

Reference:

Christopher, M., & Towill, D. (2001). An integrated model for the design of agile supply chains International Journal of Physical Distribution & Logistics Management, 31 (4), 235-246 DOI: 10.1108/09600030110394914

Rate This:

Tags:

Effective Demand Forecasting and Improvements Strategies for Supply Chain Planning

Daniel Dumke — Mon, 14 Jun 2010 04:05:00 +0000

The article reviewed here takes a look at typical biases in supply chain demand planning and how to avoid it. This work could prove very valuable for many companies who rely on manually adjusted forecasts.
Usually the forecasting process uses two steps:
1) statistical forecast by the forecasting system
2) manual adjustment to include additional effects (eg. additional analysis of demand pattern not included in step 1)

Data

The authors acquired data by four supply chain companies (industries: pharmaceuticals, food, household products and one retailer). As a whole over 68’000 datasets (statistical forecast, final (adjusted) forecast, actual outcome) were analyzed.

Problem

Now the problem seams to be that the human forecaster has several biases (systematic deviations from the real forecast). Eg. so that usually the human forecast is too positive.

Solution

Three mitigation strategies are suggested:

Blattberg-Hoch approach
using the inputs of the human forecaster and the model each with a 50% share of the whole forecast.
Error bootstrap rules
The goal here is to include knowledge about prior errors in the forecasting process
Avoiding small adjustments
as the results show small adjustments do on average not improve the overall forecast

Reference:

Fildes, R., Goodwin, P., Lawrence, M., & Nikolopoulos, K. (2009). Effective forecasting and judgmental adjustments: an empirical evaluation and strategies for improvement in supply-chain planning International Journal of Forecasting, 25 (1), 3-23 DOI: 10.1016/j.ijforecast.2008.11.010

Rate This:

Tags:

demand uncertainty

forecast

risk

Modeling uncertain forecast accuracy in supply chains with postponement

Daniel Dumke — Fri, 16 Apr 2010 08:24:03 +0000

LeBlanc, Hill and Harder emphasis the uncertainty in forecast accuracy in their 2009 paper. And therefore address a huge gap in current SCRM research.

Model

In the model of LeBlanc et al. exist two uncertainties: Uncertainty over forecast accuracy and demand uncertainty.

Source: LeBlanc et al. (2009)

There are two decision points where the manager can decide if / how much he wants to produce: A and B. At time C the finished goods are shipped. Restrictions to the production quantity at A an B can be applied.

LeBlanc et al. address three questions with this model.

How much should the firm pay to improve the accuracy of the forecast?
How much should the firm pay to achieve strategic changes, such as reducing the cost of incurring each shortage or the cost of delaying procurement and production until time B?
For different forecast accuracies, shortage and holding costs, postponement add-on percentages, etc., what percentage of the forecast should a manufacturer postpone until time B, instead of producing at time A?

Solution

The solution is done in using dynamic programming. The random variables have overlapping sums, the Forecast is calculated by FC = X_1 + X_2; the actual demand is calculated by AD = X_2 + X_3.
Using this technique it is easy to simulate different forecast accuracies: If X_2 is low compared to X_1 and X_3, forecast accuracy is low as well, and vice-versa.

Result

After implementing this model in a computer, it is possible to answer the above mentioned questions, and especially quantify the benefit of better forecasting accuracy.

Reference:

LeBlanc, L. J., Hill, J. A., & Harder, J. (2009). Modeling uncertain Forecast Accuracy in Supply Chains with Postponement Journal of Business Logistics, 30 (1), 19-31

Rate This:

Tags: