Establishing a demand forecasting tool that helps control your company's product inventory
The Analytical Approach — Combining Forecasting and Optimization
One of the biggest challenges for companies from different sectors of the economy is the control of the inventory levels of their products, both in the final stages and in intermediate stages of the chain. The greater the added value of the products, the greater the opportunity cost of maintaining a high level of inventory to ensure that there is no break in the chain, however, in cases of high product movement, significantly reducing inventory levels makes the operation extremely vulnerable to disruption events. In this scenario, using an analytical approach, based on an accurate demand forecast and an optimization model, to define when to insert and remove a particular product from the chain is the most recommended. Thus, it becomes a more efficient way to minimize operational, logistical and financial costs, in addition to controlling the trade-off between service level, logistics costs, and financial costs.
It is essential that from the beginning of problem modeling, stakeholders are involved in the discussion, construction, and validation of the model
The implementation of analytical modeling for the management of high value-added inventories must pay attention to operational details to ensure that the results obtained in simulations are in fact achieved. It is essential that from the beginning of the modeling of the problem, the particularities of the operation are considered and the stakeholders are involved in the process of discussing, constructing and validating the model. In this way, the details of the operation will be translated into mathematical modeling, even treating special cases and ensuring that the areas will have full confidence in the model for decision-making.
In the analysis of inventory management in a chain, there are two factors that are susceptible to high variability: the demand for products and the time needed for those products to be available. The use of a technically-based methodology brings a series of benefits to the operation, by:
- Ensure the supply of products with a service level appropriate to the operation
- Balance the total cost of the chain, seeking global optimal scenarios
- Plan inventory movements properly, giving visibility of product exits and entries at all links in the chain
Demand forecasting — the first step to good inventory planning
The construction of a model begins with an accurate demand forecast. Achieving accurate inventory flow, minimizing errors and identifying the variables that actually impact the variation in demand for a product in the chain, is the first step for optimizing volumes at each stage of the value chain.
To develop a demand forecasting model, it is initially necessary to understand the nature of the operation to assess what are the main factors that influence its variation over time. After this understanding, it is possible to evaluate which methods are best suited to the particularities of the operation.
Among the possible options, it is possible to separate them into two categories: temporal methods and causal methods.
Demand forecasting methods
Classical Decomposition: Method for predicting statistics that consists of the decomposition of a historical series into the components of trend, seasonality, and cycle for the projection of these values into the future.
Exponential Damping: As with classical decomposition, this method decomposes the series into trend, seasonality, and cycle components for the projection of values for the future, however, it is possible to differentiate the importance of the most recent data from the older ones.
Multiple Regression: A causal predictive method that consists of using independent variables to learn about their relationship with a dependent variable. This is one of the simplest methods in the machine learning technique (Machine learning).
Neural Networks: Machine learning technique that consists of the use of several variables to constitute layers of relationships between variables, simulating a brain's neuron network and its connections.
The first is based on analyzing demand data in a chronological manner and applying it to predict future values. This category includes, for example, the following methods: Classical Decomposition and Exponential Damping.
The second is based on the use of variables that explain demand and, with the projection of these variables, the forecast is constructed. Multiple regression and neural networks are examples of causal methods. This type of method has the advantage of allowing the inclusion of variables to explain different seasonalities, such as day of the week, business day of the month, week of the month, in addition to external factors such as inflation and holidays. The main difficulty in using causal methods is the need to predict future values for all of these variables.
Using the demand forecasting methods mentioned above, it is possible to create a tool that tests each of the methods, determines which of them has the highest accuracy, and applies this method to demand forecasting. This tool is applicable in various businesses, operations, or industries, such as forecasting the inflow and outflow of money at bank branches and cash machines, products and support services carried out by backoffice areas, and the sale of products from a retail gift company.
Managing inventory levels — the trade-off between overstock and stockout risk
After obtaining a demand forecast, it is necessary to define the strategy adopted for inventory planning. Considering high value-added inventories, it is essential to have a complete view of the costs and factors that impact the chain, such as:
- Cost of transporting products
- Opportunity cost associated with inventory volume
- Cost of receiving and shipping the product in each shipment
- Safety risks associated with transporting the products
- Safety risks associated with maintaining a very high inventory level
- Operation service level
Once the inventory management model has been defined, the implementation must be carried out with close attention to the details of the operation, with the team involved in daily life feeling part of the new model. The user must recognize the final value of the work and ensure alignment between all stakeholders, avoiding conflicts of interest between the areas impacted by the operationalization of the model.
Because it is a process with the potential to involve significant changes in the operation, change management is essential. The transition to a new decision-making model is a critical process, interfering with the routine of areas that are impacted by inventory levels and several employees, who must be trained to operate the deployed model.
The factors that lead to successful deployment
With the change in the company's mindset caused by changes in inventory management policy based on a mathematical model that minimizes costs globally, some stages of the chain may start to cost a higher amount. If the company opts for a model that reduces the volume of inventories, a higher cost of transportation may be required, but this will be offset by a reduction in opportunity cost, impacting the reduction of the overall cost of the operation. For this transition in the operation's mentality to be carried out successfully, it is necessary to ensure three fundamental points: goal alignment, team empowerment, and model accuracy.
- Goal alignment — The guarantee that everyone will be pursuing the same objectives: The new inventory management model will not have its maximum potential obtained if there is a misalignment between the areas impacted by the operation. To mitigate this risk, the operation's goals must be defined in accordance with the inventory management model, to avoid that, for example, a stakeholder's goal is to minimize the logistics cost while the new model increases the logistics cost to bring a greater return in opportunity cost.
- Team training — The guarantee that the Inventory Planning area is aligned with the new policies: The change of mindset in the inventory policy will raise questions from various areas of the company. For example, updating this policy may work with lower security levels, which may lead to an increased occurrence of system disruption. For this idea to be disseminated, it is necessary that the main stakeholders in the area responsible for inventory planning are fully trained to answer questions and transmit a positive view of the model to the other focal points affected by the change.
- Accuracy of the model: The guarantee of effectiveness in the face of situations commonly faced in the management of company inventories: In the implementation of a new inventory planning model, specificities will emerge that may not have been mapped at the time of developing the mathematical modeling that solved the problem. It is essential that these situations be analyzed and the model adapted to meet the peculiarities of the operation, otherwise important focal points may start to use exceptions as a rule and discredit mathematical modeling. For this reason, it is important to transfer knowledge and skill development to the team responsible for the operation and maintenance of the model, ensuring that adjustments and evolutions are carried out with high responsiveness.
Having noted all the points above, the implementation of a robust model, aimed at minimizing the overall cost of the operation, has enormous potential to generate results, especially in scenarios of high interest rates and the consequent increase in the cost of capital linked to the company's inventory. The ability to manage the operation, combined with analytical expertise for the development and operation of a forecasting and optimization model, is the key for the Inventory Planning process to be reviewed and for an organic and sustainable evolution.
About the authors
Marcus Sousa is a Visagio consultant, a specialist in projects focusing on management models, supplies and analytics in the retail, financial market, metallurgical industry and energy sectors. He also acts as the leader of Visagio's Research & Intelligence area, focusing on market research and analysis.
Felipe Pena is a Visagio consultant, a specialist in projects focusing on budget management, process engineering and analytics in the purchasing and banking sectors.