Greenfield Discovery

In this use case, we demonstrate how to use Convect AI’s supply chain network design tool to solve a Greenfield Discovery problem, where the goal is to determine optimal locations for new warehouses to serve a set of customers efficiently. This is particularly useful when entering a new market or expanding operations into an unserved region.

Problem Statement

We have 100 customers distributed across the state of Virginia, each with varying demand levels. To serve these customers, we need to build between 3 to 5 warehouses. Our objective is to meet customer demand while adhering to several key constraints:

1. Service Distance: Each customer should have at least one warehouse within 15 km.
2. Service Level: At least 95% of the total customer base should be served by a warehouse within the 15 km service distance.
3. Warehouse Cost: Each warehouse has a fixed operational cost of $500,000.
4. Warehouse Capacity: Each warehouse has a maximum capacity, which limits the total customer demand it can fulfill.

Using these constraints, we aim to answer the following critical questions:

1. What is the optimal number of warehouses to build (3, 4, or 5) to balance cost and service?
2. Where should the warehouses be located to best serve the customer base?
3. Which customers will each warehouse serve, ensuring that demand is met while minimizing transportation costs?

The heatmap shows customer locations with color intensity representing the demand level.

Preparing Data

Before solving the Greenfield Discovery problem, it's crucial to prepare the necessary input data. The spreadsheet file contains three key sheets: Parameters, Customers, and Sites. Let’s go over the content of each sheet and how to use it to configure Convect AI’s network design tool.

Download Sample File

1. Parameters Sheet:

This sheet contains several key configurations that define the constraints and goals of the warehouse placement optimization. The most important parameters for this use case are:

• Minimum Number of Sites: 3 – The minimum number of warehouses to be built.
• Maximum Number of Sites: 5 – The maximum number of warehouses to be built.
• Minimum Site Capacity: 1000 units – The minimum capacity each warehouse should handle.
• Maximum Site Capacity: 30,000 units – The maximum capacity each warehouse can handle.
• Maximum Service Distance: 15 km – Every customer must have at least one warehouse within 15 km.
• Minimum Service Coverage Percentage: 95% – At least 95% of all customers must be served by a warehouse within the 15 km service distance.
• Default Site Cost: $500,000 – The fixed operational cost for each warehouse built.

These parameters define the model's operational constraints, ensuring that the solution meets both cost efficiency and service-level requirements.

2. Customers Sheet:

This sheet provides detailed information about the 100 customers in Virginia, each with varying demand levels. The key columns include:

• Customer ID: A unique identifier for each customer.
• Customer Address: The location of the customer.
• Latitude and Longitude: Geographical coordinates, which will be used to calculate distances between customers and potential warehouse sites.
• Customer Demand: The total demand from each customer, which must be met by one or more warehouses.

This data will be used to assign customers to the closest warehouse locations and ensure their demands are fulfilled.

3. Sites Sheet:

This sheet is currently empty, which means that the software will determine optimal warehouse locations based on the customer data and the constraints provided in the Parameters sheet.

Using the Network Design Tool

Follow these steps to use Convect AI's network design tool to solve the Greenfield Discovery problem:

1. Login to Convect AI's Flow Platform Go to https://flow.convect.ai and log in.

2. Open the Network Design App Select the Network Design app from the platform's main page.

3. Create a New File and Upload Input Data Click the "Create File" button, enter any required information, and upload your prepared input data file containing the Parameters and Customers sheets. Wait for the data import to finish.

4. Solve the Problem Click the Solve button and wait for the optimization process to complete.

5. View Results Check the Output View tab for spreadsheet results or use the Graph View tab for visualizations of the warehouse locations and customer assignments.

By following these steps, you'll get the optimal number and location of warehouses and see which customers each warehouse will serve.

Analyzing Results: Comparing the Scenarios

After running Convect AI's network design tool, three scenarios—ck-3 (3 warehouses), ck-4 (4 warehouses), and ck-5 (5 warehouses)—were generated. Let’s compare them to answer our key questions:

1. What is the optimal number of warehouses?
2. Where should the warehouses be located?
3. Which customers will each warehouse serve?

1. What is the Optimal Number of Warehouses?

The following table summarizes the key metrics for each scenario:

Scenario	Number of Warehouses	Service Level	Total Cost	Total Distance (km)	Customers Served (%)
ck-3	3	96%	$5.66M	8,439 km	96%
ck-4	4	100%	$5.43M	6,774 km	100%
ck-5	5	100%	$5.44M	5,953 km	100%

• Scenario ck-4 (4 warehouses) provides the best balance between cost and efficiency. It achieves 100% customer service coverage with the lowest total cost ($5.43M) and a relatively low total travel distance (6,774 km).
• While Scenario ck-5 (5 warehouses) offers a marginal improvement in travel distance (5,953 km), it comes with a slightly higher cost ($5.44M), making it less optimal compared to ck-4.

2. Where Should the Warehouses be Located?

For each scenario, the tool determines the optimal locations of warehouses. Below are the warehouse locations for Scenario ck-4, as shown in the map visualization:

Warehouse	Latitude	Longitude	Assigned Customers	Assigned Capacity
Warehouse 1	37.02	-81.74	15	5,358
Warehouse 2	37.21	-76.77	34	17,774
Warehouse 3	37.46	-79.28	29	14,814
Warehouse 4	37.82	-77.17	22	10,517

Below is the map for Scenario ck-4 showing customer-site assignments:

All customers are within the service range of a warehouse.

For comparison, Scenario ck-3 (3 warehouses) places customers farther from their assigned warehouse, resulting in a lower service level and higher travel distances:

Four customers (in red) fall outside the service range of any warehouse, resulting in a 96% service level.

On the other hand, Scenario ck-5 (5 warehouses) reduces travel distances further but slightly increases costs:

A denser network of warehouses reduces overall travel distances but slightly increases the total cost compared to the 4-warehouse scenario.

3. Which Customers Will Each Warehouse Serve?

In Scenario ck-4, the Customer-Site Assignment data provides details on which warehouse serves each customer. For instance:

• Warehouse 1 serves 15 customers with an average distance of 51 km.
• Warehouse 2 serves 34 customers with an average distance of 74 km.

These assignments are optimized to meet the service level constraints while minimizing the distance between customers and warehouses. The tool assigns each customer to the closest available warehouse that can meet their demand while adhering to the 15 km service distance requirement for 95% of customers.

Conclusion

• Optimal Number of Warehouses: The results suggest that Scenario ck-4 (4 warehouses) is the best solution, offering 100% customer service coverage at the lowest cost ($5.43M) and a reasonable total distance (6,774 km).
• Warehouse Locations: Refer to the Sites Found sheet and the map visualizations to see the exact warehouse locations for each scenario.
• Customer Assignments: Use the Customer-Site Assignment sheet to analyze which customers are assigned to each warehouse and the distance between them.

In conclusion, Scenario ck-4 strikes the best balance between cost, service level, and total travel distance, making it the recommended solution for the Greenfield Discovery problem.