Reduce costs by improving inventory planning and optimizing the stockpile through Machine Learning and mathematical optimization.

Discover how a leading company in Energy & Utilities improved the inventory planning and reduced working capital costs with no decrease in the level of service.

What-if analysis to reduce working capital costs without losing the quality of the offered services.

One of the main players in the Energy & Utilities sector needed a reduction of the costs associated with in-warehouse working capital, yet with no decrease in the quality of the service provided.

So, aHead Research proposed the creation of a mathematical model that represented the stockpile in a warehouse of replacement parts, with the goal of optimizing the quantities of the items and consequently reduce the working capital and warehousing costs.
In particular, we were asked to identify the best restocking and replacement policies for the parts present in such warehouse. The strategy to adopt could not be simplistic, as it had to consider several optimization parameters such as restocking quantities and reorder points in the case of restocking strategies, or when to restock -and how much- in case of on-demand strategies. In order to consider the entirety of such factors, a what-if analysis was necessary to show the results relating to the various strategies and the services offered. Alongside the analysis, a simulation that could show and allow for the evaluation of the proposed solutions was also needed.

Key questions:

What kind of restocking strategy allows for the optimization of warehousing costs?
In the case of a re-stocking strategy, when is it necessary to re-stock and what is the correct reorder point?
In the case of an on-demand strategy, when -and how much- is it necessary to restock?
If a safety stock is employed, what should its size be in order to offer a certain level of service without prohibitive costs? Can a safety stock be avoided?
Which items require a specific restocking strategy?

To answer such questions, advanced algorithms provide support to decision-making processes by considering all pertinent factors. Increasing the size of the stockpile can lead to a better saturation of the transport fleet while reducing fixed administrative costs of ordering, and decreasing movement in the warehouse, which could be problematic in case of costly and time-demanding set-ups. Nevertheless, high stockpile quantities equal high warehousing costs, including purchase of the items, maintenance, obsolescence, and fiscal/insurance costs.

our method

THE PROBLEM

THE SOLUTION

THE BENEFITS

The problem

Reducing costs of the working capital by optimizing the stock in the warehouse

Choosing what policy to apply to the warehouse stockpile is a difficult decision for managers, as there may be hundreds of thousands of items to sort out. The wider the array of products on the market, the greater the difficulty in adapting the stock to market demand. Each product presents specific critical issues related to its features, its lead-time, and its costs. Immediate availability of materials, components or end products is essential to satisfy customer requests, both in terms of minimal quantities for quick delivery and to be able to face emergencies. An optimal sizing of the warehouse allows for higher elasticity and flexibility in terms of the kind of items produced as well as the quantities. The trade-off here is clear: it is necessary to reduce the quantities of the stockpile as much as possible in order to reduce costs associated with the working capital, yet a too drastic reduction might lead to extremely high costs related to the handling of issues caused by the lack of products in demand.

The solution

Our solution for inventory planning and stockpile optimization in the warehouse

The three team of Artificial Intelligence, Math-Optimization, and Simulation of aHead Research developed a Python-based solution that incorporates a machine-learning model for the classification of products in the stockpile and a Simulation-based Optimization model able to pinpoint the best restocking parameters for any adopted strategy. Such technology made it possible to shrink planning time by managing thousands of different products at the same time.

The adopted algorithm can cover all operations needed for a correct planning, such as:

Classifying products, components and materials based on how much they are employed and any technical issue;
Automatically associating the best reordering policy and choosing correct parameters to apply to it;
Simulating different scenarios and calculating the KPIs necessary to evaluate them;
Managing any potential lack in data input.

Taking an informed decision requires classification, optimization, what-if analysis, and simulation that only a sophisticated AI-based software can provide. A powerful tool like Machine Learning allows for the classification of the products present in the warehouse according to potential issues and degree of usage starting from basic information such as the characteristics of the products and KPIs such as lead-time or number of movements, both inbound and outbound. Together with simulation, Math-Optimization identifies the correct parameters of the chosen strategy.

The benefits

The benefits gained from inventory planning and stockpile optimization

The proposed solution allowed for the reduction of working capital costs associate with the client’s warehouse by 47%, without considering the cost reduction connected with storing and reordering. Th analysis made it possible to compare past data with the different reorder strategies and the optimized parameters. The KPIs identified a sharp reduction of the final stockpile by 28% and an 11% increase in turnover rate, meaning that the best strategy can lead to shorter turnover periods with less stock in the warehouse. Although the average safety stock was lowered by 12%, the overall number of occasions in which the safety stock was needed was reduced by 55%, therefore improving the overall resilience to unpredictable spikes in demand.

The combination of mathematical optimization, artificial intelligence and simulation made it possible to achieved outstanding economic results, i.e., -47% of overall working capital costs in view of a 2-year planning without losing the quality of service.