Executive Briefings

Stochastic Transportation Modeling

Most large shippers spend a great amount of time and expense collecting, analyzing and maintaining the data used to drive daily transportation planning and execution.  I refer to these data, when codified and integrated into the shipper's transportation system, as the organization's transportation policy.  The policy is composed of lanes, modal options, rates, carrier and fleet capacity and service levels as well as a multitude of other decision variables and rules that must work in concert to drive daily decision making.

As the transportation policy defines the framework with which all transportation decisions will be made, it is imperative that it is constructed in a manner that is not overly constricting.  It needs to be flexible enough to provide the advanced optimization capabilities of the transportation management system the latitude to find savings.  It also needs to be built in a manner that is cognizant of typical variability that is found within virtually every supply chain.  It is on this point that I feel there are opportunities for improvement over existing processes.

Deterministic vs. Stochastic Strategic Planning

When developing the data that make up the transportation policy, most analysts still rely heavily on statistical averages.  This approach completely ignores the inherent variability that underlies nearly every aspect of transportation.  From order sizes, to lane volumes, to travel times to fuel prices, using a single, discrete value to represent each of these variables during strategic planning will lead to a poorly designed transportation policy that will ultimately be reflected as inefficiencies in daily planning.

So why do most organizations develop their policy in this manner?  The simple answer is that averaging is both easy to comprehend and calculate.  Additionally, averages also provide a discrete, numeric value than can be used by deterministic optimization tools that are sometimes used in transportation modeling.  Unfortunately, averages rarely reflect reality.   For example, if half the time a supplier ships 5,000 pounds and the other half they ship 40,000 pounds, then the average is 22,500 pounds. But in reality the supplier never ships a 22,500-pound orders.  Therefore, a fundamental input used as the basis for the development of the transportation policy is completely wrong.

To be clear, when doing daily planning, using a deterministic planning tool is usually (there are examples in long-haul networks where stochastic planning is applicable) the most logical method since order quantities, rates, capacity and other values are known.  However, strategic planning should utilize a "stochastic" approach that is based on calculated probabilities, not the oxymoronic concept of forecasted certainties that is implied in the deterministic approach.  Applying stochastic principles is not new to supply chain planning. The concept of safety stock exists precisely due to the realization that we cannot predict with any certainty customer demand, lead times or order fill rates.  The latest transportation modeling technology allows for combined stochastic optimization and simulation, which enables analysts to incorporate known variability into the process of developing the transportation policy.

So, how does this all work in the real world?

Let's take the example of prepaid-to-collect conversion. We want to know which should be prepaid and which should be collect.

First, this decision is driven by the difference between the freight allowance provided by the supplier and the transport cost that would be incurred by the customer. Although there are many constants in this example that might lead one to a deterministic solution, prepaid to collect is a long-term, strategic decision that is subject to significant levels of variability that must be taken into account. Key elements that need to be modeled for variability include:

• Order quantities

•  Order quantities of neighboring suppliers (i.e., Can LTL shipments be converted into multi-stop TL movements?) This is important, because looking at each supplier independently will not locate the benefits associated with inbound multi-stop TL consolidation.

•  Freight rates

•   Fuel prices. If no fuel surcharge is charged back to the supplier, the customer is taking all the risk around fuel price volatility.

•  Order frequency

•  Transport lead times. Do I have time to perform consolidations or are the ordering patterns such that I have to regularly use direct and/or expedited freight?

A deterministic approach would be to take an average ship quantity for each of the suppliers and optimize it through a deterministic solver.   However, that would assume that you understood the rate and fuel surcharge that would be used through-out the year and the order quantity would vary little.  Through stochastic optimization and simulation, the variability of these data can be modeled in a manner that provides the analyst a greater degree of confidence that the "best" decision as defined not by a snapshot of data, but over time will be made.

In summary, most organizations today take a deterministic approach when developing their transportation policy simply because it's easy to do and it's the way it has always been done.  However, long-term strategic planning must take into account known variability.  Only through taking into account variability will analysts be able to create an operationally resilient and efficient transportation policy that reflects the reality of an ever-changing world.

Source: Manhattan Associates

 

Most large shippers spend a great amount of time and expense collecting, analyzing and maintaining the data used to drive daily transportation planning and execution.  I refer to these data, when codified and integrated into the shipper's transportation system, as the organization's transportation policy.  The policy is composed of lanes, modal options, rates, carrier and fleet capacity and service levels as well as a multitude of other decision variables and rules that must work in concert to drive daily decision making.

As the transportation policy defines the framework with which all transportation decisions will be made, it is imperative that it is constructed in a manner that is not overly constricting.  It needs to be flexible enough to provide the advanced optimization capabilities of the transportation management system the latitude to find savings.  It also needs to be built in a manner that is cognizant of typical variability that is found within virtually every supply chain.  It is on this point that I feel there are opportunities for improvement over existing processes.

Deterministic vs. Stochastic Strategic Planning

When developing the data that make up the transportation policy, most analysts still rely heavily on statistical averages.  This approach completely ignores the inherent variability that underlies nearly every aspect of transportation.  From order sizes, to lane volumes, to travel times to fuel prices, using a single, discrete value to represent each of these variables during strategic planning will lead to a poorly designed transportation policy that will ultimately be reflected as inefficiencies in daily planning.

So why do most organizations develop their policy in this manner?  The simple answer is that averaging is both easy to comprehend and calculate.  Additionally, averages also provide a discrete, numeric value than can be used by deterministic optimization tools that are sometimes used in transportation modeling.  Unfortunately, averages rarely reflect reality.   For example, if half the time a supplier ships 5,000 pounds and the other half they ship 40,000 pounds, then the average is 22,500 pounds. But in reality the supplier never ships a 22,500-pound orders.  Therefore, a fundamental input used as the basis for the development of the transportation policy is completely wrong.

To be clear, when doing daily planning, using a deterministic planning tool is usually (there are examples in long-haul networks where stochastic planning is applicable) the most logical method since order quantities, rates, capacity and other values are known.  However, strategic planning should utilize a "stochastic" approach that is based on calculated probabilities, not the oxymoronic concept of forecasted certainties that is implied in the deterministic approach.  Applying stochastic principles is not new to supply chain planning. The concept of safety stock exists precisely due to the realization that we cannot predict with any certainty customer demand, lead times or order fill rates.  The latest transportation modeling technology allows for combined stochastic optimization and simulation, which enables analysts to incorporate known variability into the process of developing the transportation policy.

So, how does this all work in the real world?

Let's take the example of prepaid-to-collect conversion. We want to know which should be prepaid and which should be collect.

First, this decision is driven by the difference between the freight allowance provided by the supplier and the transport cost that would be incurred by the customer. Although there are many constants in this example that might lead one to a deterministic solution, prepaid to collect is a long-term, strategic decision that is subject to significant levels of variability that must be taken into account. Key elements that need to be modeled for variability include:

• Order quantities

•  Order quantities of neighboring suppliers (i.e., Can LTL shipments be converted into multi-stop TL movements?) This is important, because looking at each supplier independently will not locate the benefits associated with inbound multi-stop TL consolidation.

•  Freight rates

•   Fuel prices. If no fuel surcharge is charged back to the supplier, the customer is taking all the risk around fuel price volatility.

•  Order frequency

•  Transport lead times. Do I have time to perform consolidations or are the ordering patterns such that I have to regularly use direct and/or expedited freight?

A deterministic approach would be to take an average ship quantity for each of the suppliers and optimize it through a deterministic solver.   However, that would assume that you understood the rate and fuel surcharge that would be used through-out the year and the order quantity would vary little.  Through stochastic optimization and simulation, the variability of these data can be modeled in a manner that provides the analyst a greater degree of confidence that the "best" decision as defined not by a snapshot of data, but over time will be made.

In summary, most organizations today take a deterministic approach when developing their transportation policy simply because it's easy to do and it's the way it has always been done.  However, long-term strategic planning must take into account known variability.  Only through taking into account variability will analysts be able to create an operationally resilient and efficient transportation policy that reflects the reality of an ever-changing world.

Source: Manhattan Associates