How to Formulate and Solve Transportation Models in Operations Research

Every agribusiness decision involving the movement of goods – from a grain elevator shipping to feed mills, or a fertilizer company distributing to regional depots – has one thing in common: transportation costs matter. Getting products from Point A to Point B at the lowest possible cost, without leaving demand unmet or supply stranded, is exactly what the transportation model in operations research is designed to solve. It is a structured, mathematical approach that turns a complex logistics puzzle into a solvable matrix problem.

What is the transportation model? 🔗

The transportation model is a special class of linear programming problem. Its objective is to transport a single commodity from various origins (such as farms, factories, or warehouses) to different destinations (such as markets, retail outlets, or processing plants) at the minimum total shipping cost. Unlike the general simplex method used for standard linear programming, the transportation problem has a unique structure that demands its own specialized solution methods.

The five key elements of any transportation problem are: sources (supply points), destinations (demand points), supply quantities at each source, demand quantities at each destination, and unit transportation costs between every source-destination pair. The goal is to determine how many units to ship from each source to each destination so that total cost is minimized, all supply is used, and all demand is satisfied.

Formulating the transportation model 🔗

The formulation begins by organizing all available information into a transportation matrix (also called a transportation tableau). This matrix presents supply origins, demand destinations, and transportation costs in a structured table format, making it easy to see the complete picture at once. Rows represent the sources and columns represent the destinations. Each cell in the matrix holds the cost of shipping one unit from a particular source to a particular destination.

Consider a practical example: a fertilizer company has three production plants – P1, P2, and P3 – supplying three regional distribution centers – D1, D2, and D3. Each plant has a fixed output capacity (supply), each distribution center has a fixed requirement (demand), and the cost of shipping one tonne between any plant-center pair varies based on distance and logistics. These costs are entered into the matrix cells.

Balanced vs. unbalanced problems 🔗

A transportation problem is called balanced when total supply equals total demand. When the sum of supply from all sources equals the sum of demand across all destinations, the problem is balanced. When they don’t match, the problem is unbalanced. To handle this, a dummy source or dummy destination is added with a supply or demand value equal to the difference, and transportation costs for the dummy row or column are set to zero. This converts the unbalanced problem into a balanced one before proceeding to find a solution.

Mathematical objective function 🔗

Once the tableau is set up, the problem is expressed mathematically. Let x_ij represent the number of units shipped from source i to destination j, and c_ij represent the cost per unit for that route. The objective function is the total transportation cost, expressed as a minimization of the sum of all c_ij × x_ij values. The constraints require that shipments from each source do not exceed its supply, and that shipments to each destination meet its demand exactly.

Finding the initial feasible solution 🔗

Solving a transportation problem involves two phases: first finding an initial basic feasible solution, then testing and improving it for optimality. Three standard methods are used to find that initial solution.

North-West Corner Rule (NWCR) 🔗

This is the simplest method to apply. The process starts at the top-left (north-west) cell of the tableau and allocates as many units as possible – limited by either the source’s supply or the destination’s demand – before moving right or down. The allocation continues cell by cell until all supply and demand values are exhausted.

The main advantage is speed and simplicity. The significant drawback is that the NWCR ignores transportation costs entirely during allocation, which often produces a starting solution that is far from optimal and requires more iterations to improve.

Least Cost Method (LCM) 🔗

The Least Cost Method directly addresses the main weakness of the NWCR by factoring in costs from the start. Allocations begin with the cell that has the lowest unit transportation cost, with the maximum possible quantity assigned to that cell before moving to the next cheapest available cell. This produces a better initial solution than the NWCR because it prioritizes low-cost routes first.

In practice, for agribusiness applications – such as determining which farm should supply which processing plant – the LCM is more practically useful than NWCR because it already reflects economic logic: use the cheapest routes first.

Vogel’s Approximation Method (VAM) 🔗

VAM is the most sophisticated of the three initial solution methods and consistently produces the best starting point. VAM works by calculating a penalty cost for each row and column – the difference between the two lowest unit costs in that row or column. This penalty represents the extra cost incurred if the cheapest route in that row or column is not used.

The steps of VAM are straightforward: calculate penalties for all rows and columns, select the row or column with the highest penalty, allocate as much as possible to the cell with the lowest cost in that row or column, eliminate the satisfied row or column, and repeat the process until all allocations are made.

VAM generally produces an initial solution that is closer to optimal – or is itself optimal – compared to the NWCR and LCM, though its calculations take somewhat longer. For agribusiness managers dealing with multiple sources and destinations, the extra calculation effort upfront often saves time overall because fewer optimization iterations are needed.

Checking feasibility: the m + n − 1 rule 🔗

After obtaining an initial solution using any of the three methods, you must verify that it is a basic feasible solution before testing it for optimality. The number of allocations in the solution must equal m + n − 1, where m is the number of rows (sources) and n is the number of columns (destinations).

For example, in a 3×4 matrix (3 sources, 4 destinations), the solution must have exactly 3 + 4 − 1 = 6 allocated cells. If there are fewer, the solution is said to be degenerate, and a very small quantity (represented as ε or epsilon) is assigned to an independent empty cell to restore the required count. This is necessary to proceed with optimality testing.

Testing and optimizing the solution 🔗

An initial feasible solution is rarely the optimal one. Two standard methods are used to test and improve it.

Stepping Stone Method 🔗

This method evaluates each unused (empty) cell in the tableau by tracing a closed loop through currently allocated cells – moving only horizontally and vertically – and calculating the net change in total cost if one unit were to be shifted to that empty cell. If any unused cell shows a negative improvement index, shifting allocation to it will reduce total cost, so the reallocation is made. The process repeats until no negative improvement indices remain, confirming optimality.

MODI Method (Modified Distribution / UV Method) 🔗

The MODI method is more efficient than the Stepping Stone approach. It assigns row values (u_i) and column values (v_j) to the allocated cells using the condition c_ij = u_i + v_j, then calculates a penalty d_ij = c_ij − (u_i + v_j) for each unoccupied cell. If all d_ij values are zero or positive, the current solution is optimal. If any d_ij is negative, a reallocation is performed along a closed loop starting from that cell, and the process repeats.

Unlike the Stepping Stone Method – which requires tracing a separate closed path for every unoccupied cell – MODI only draws a closed loop for the cell with the most negative penalty, making it faster for large problems.

Practical applications in agribusiness 🔗

Transportation models are not just textbook exercises – they have direct, measurable applications in agricultural supply chains. The objective is always to minimize shipping costs while satisfying supply capacity and demand constraints simultaneously. In agribusiness contexts, this applies in several important ways.

Grain companies routinely use transportation models to decide which storage elevators should supply which processing or milling facilities. Fresh produce distributors apply them to route harvested products from farms to retail outlets before spoilage occurs. Livestock operations use them to determine the most cost-effective way to move animals from farms to processing plants. Research on vegetable supply chains in Sri Lanka showed that applying optimization models to route planning reduced total food miles by over 30% and cut CO₂ emissions by roughly 35% – a clear demonstration of the real-world impact these models can deliver.

Seasonal planning is another strong use case. Many crops have a defined harvest window, but processing and market demand continue year-round. A transportation model can be run across different time periods and storage scenarios to determine the lowest-cost distribution pattern at each stage of the season.

Handling real-world complexity 🔗

Real agricultural supply chains involve variables that the standard transportation model simplifies. Cost minimization in agricultural logistics is subject to transport routes, distance, and total transit time, and in practice, factors like road conditions, vehicle availability, and perishability timelines also come into play. Weather variability, unpredictable yields, and fluctuating market demand mean that supply and demand values are rarely as fixed as the model assumes.

The practical response is to run multiple scenarios – varying supply and demand inputs – to understand how sensitive the optimal solution is to those assumptions. More advanced formulations, such as multi-objective or stochastic transportation models, can incorporate uncertainty directly, though they add mathematical complexity. For most agribusiness planning purposes, a well-constructed standard transportation model with scenario testing provides valuable, actionable guidance.

Step-by-step summary of the solution process 🔗

To summarize the complete process for solving a transportation model: first, check balance – if total supply ≠ total demand, add a dummy row or column with zero costs. Second, find an initial feasible solution using NWCR, LCM, or VAM – VAM is recommended for closest-to-optimal results. Third, verify feasibility – confirm that the number of allocated cells equals m + n − 1; resolve degeneracy with ε if needed. Fourth, test for optimality using the Stepping Stone or MODI method. Fifth, if the solution is not yet optimal, reallocate along the identified closed loop and repeat the optimality test. Continue until all unoccupied cell penalties are non-negative – at that point, the optimal solution has been reached and the minimum total transportation cost has been found.

What do you think? If you were managing the distribution of perishable produce from multiple farms to several regional markets, which initial solution method – NWCR, LCM, or VAM – would you prioritize, and why? And how would you modify the standard transportation model to account for the unpredictability of harvest yields in your planning?

References

1. https://www.researchgate.net/publication/356469857_Transportation_Model
2. https://ijsra.net/sites/default/files/IJSRA-2024-1406.pdf
3. https://www.ijraset.com/research-paper/comparison-of-transportation-problem-in-operation-research
4. https://www.mmmut.ac.in/News_content/11235dep-notice_11072020.pdf
5. https://egyankosh.ac.in/bitstream/123456789/20787/1/Unit-4.pdf
6. https://operationresearchandrada.wordpress.com/transportation-model/
7. https://maxusknowledge.com/lectures-on-transportation-model/
8. https://www.uobabylon.edu.iq/eprints/publication_3_29154_132.pdf
9. https://www.geeksforgeeks.org/dsa/transportation-problem-set-6-modi-method-uv-method/
10. https://www.slideshare.net/slideshow/modisteppingstonepptx/257183350
11. https://www.mygreatlearning.com/blog/transportation-problem-explained/
12. https://www.mdpi.com/2071-1050/16/17/7601
13. https://www.tandfonline.com/doi/full/10.1080/13675567.2022.2107625