Data Science to Implement a Circular Economy | by Samir Saci | Mar, 2024

You have worked with the logistics teams for years to implement reporting and optimization tools for forward logistics.

To prepare for the transition to a circular economy, they have additional requests to manage collection and sorting processes.

What are the additional processes that we need to monitor and optimize?

In this section, we will briefly cover several examples of process design requirements with the associated analytics solutions.

Descriptive Analytics: Product tracking and traceability

Because of regulations, you must ensure that products are used and processed in the most sustainable way possible.

Therefore, the logistic teams need to be able to track and trace products throughout their lifecycle.

For forward logistics, products are tracked from the factory to the store

• Master data databases will include (Stock Keeping Unit) SKU Number, product information (size, colour, packaging)
• Production Systems can provide the batch number (useful if you need to call items), production date and factory location.
• Warehouse and Transportation Management can track the product along the logistics chain, from warehouse receiving to store delivery.

How can you support the tracking of returned items with advanced analytics?

For reverse logistics, the logistics request your support in monitoring the flow of “returned” products from the collection box to the sorting centre.

Different logistics systems that are recording transactional data can be used as data sources to build automated flow monitoring tools:

• Enterprise Ressource Planning (ERP) records the customer returns at the store with Store Location, SKU ID, Quantity, Collection Box ID and Collection Time.
• Transportation Management Systems records the pick-up of collection boxes, including store location, collection box ID and collection time.
• Warehouse Management Systems tracks the items from receipt to the end of the sorting process with the SKU ID, Collection Box, Receiving Time, End of Sorting Time, and Final Destination.

Business Intelligence methodologies using a central data warehouse can support the creation of harmonized tracking data sources that can be used for audits or reporting.

If you need more details on how to implement it, have a look at this article.

Now that you have provided transparency to the operations, let’s focus on process optimization.

Prescriptive Analytics: Sorting Network Design

The logistic teams would like to design a network of sorting centres that minimizes the environmental impact of the reverse flow.

The optimal network may differ from the current one as you need to collect items after sales and reintroduce recycled materials into the chain.

They shared with you

• A list of potential sorting locations with their capacity (Units/Day)
• Forecasts of reverse flow volumes from the stores (Units/Day)

Question: Where should we place the sorting centers to minimize costs and CO2 emissions?

This will remind you of the Supply Chain Network Optimization Problem for which you need to design a network of factories.

The market demand must be fulfilled by factories with limited capacities and different costs to produce and deliver goods.

Linear programming is used to select the right factories that minimize the overall costs while respecting a set of constraints.

For more details, have a look at this detailed example

How can we adapt this solution to the sorting network design problem?

This can be easily adapted to this new problem

• Demand => Units Collected from the Stores (Units/Day).
• Factories Capacity => Sorting Capacity of each Centre (Units/Day).

You can then select the optimal set of sorting locations depending on the objective you have

• Minimizing the total costs?
  The algorithm will consider the transportation costs (from the stores to each sorting location) and the sorting costs of each centre.
• Minimize the CO2 emissions?
  The algorithm will try to minimize the distance from the store to the selected centres.

Now that you have selected the optimal set of sorting centres, you can organize the reverse flow.

Prescriptive Analytics: Reverse Flow Optimization

The transportation team requests your support in designing a tool to allocate the right sorting centre to each store.

Indeed, as the volumes and capacities evolve, you’d better dynamically allocate a sorting centre to each store on a weekly or daily basis.

Question: For each store i, which centre should sort the returned items to minimize the transportation costs?

In a previous article, I tackled a similar problem: The Supply Planning Problem.

Several factories replenish distribution centres that store goods and deliver stores.

To answer this problem, we also use linear programming to

• Optimize upstream flows from the factories to the distribution centres
• Deliver each store from the right warehouse

This solution can be used after slight adaptations

1. Define (or forecast) the volumes from the sorting centres to the final recycling location
2. Define the capacity of each sorting centre

You can then follow the approach (and use the code) detailed in this article.

Now that we have designed the network of sorting locations and developed a tool to optimize reverse flows, we can work on performance monitoring.