Inventory Estimation Using Demand Share and Variance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current inventory management systems struggle to accurately estimate target inventory for items with limited data, such as new or seasonal items, using only partial information about demand distribution, and they are not scalable for online retail settings.
Innovation Solution
A system that uses limited information like mean, variance, and unimodality to estimate target inventory, focusing on desired quality of service constraints, and is designed to be computationally efficient for large-scale inventory decisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If robust optimization techniques are used to handle quality of service constraints with limited demand distribution information, then the service level guarantee is improved, but the computational complexity increases making it unsuitable for scaling to millions of items
Solution Approach 1:
The patent transforms the complex robust optimization problem into a simpler parametric form by expressing the target inventory as a function of key parameters (mean demand, variance, service level). This allows the solution to maintain reliability guarantees while becoming computationally efficient for large-scale deployment across millions of items.
Solution Approach 2:
The patent extracts and isolates the essential elements needed for inventory decision-making (mean, variance, service level) from the complex full probability distribution, creating a simplified model that retains the critical information needed for reliable inventory management without the computational burden of complete distributional assumptions.
2Measurement precision
If current inventory management systems use full probability distribution knowledge to make accurate demand predictions, then the prediction accuracy is improved, but the system becomes inapplicable to items with limited data such as new, seasonal, or slow-selling items
Solution Approach 1:
The patent applies partial action by using only the essential moments (mean and variance) of the demand distribution rather than requiring the complete probability distribution. This partial information approach provides sufficient accuracy for inventory decisions while enabling the system to handle items with limited historical data that cannot support full distributional estimation.
3Reliability
If min-max approach or robust optimization techniques are used to minimize expected cost under worst-case demand distribution, then the service level constraint is satisfied, but the cost function includes components that are hard for practitioners to quantify
Solution Approach 1:
The patent creates a simplified copy or approximation of the complex min-max optimization problem that preserves the essential service level guarantee functionality while replacing hard-to-quantify cost components with directly observable and understandable parameters like mean demand, variance, and target service level, making the system easier to operate and parameterize.
Data Source
AI summary
Systems and methods, and computer readable media for target inventory estimation of a geographical region are disclosed. The method receives an item identifier associated with an item for target inventory estimation. The method may access the demand share estimate of the item in the geographical region and the mean and variance estimates of the demand share estimate in the geographical region. The method may then access overall demand forecast data for the item and calculate the mean and variance estimates of the demand forecast. The method may use the generated mean and variance estimates and demand share estimate of the geographical region to determine the target inventory estimation of the region.