Inventory Allocation Using INF Prediction and Stock Suppression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current inventory systems lack the ability to generate accurate inventory availability data in real-time, leading to inventory not found (INF) events, which result in split orders, order delays, and underutilization of high-quality nodes, causing a negative customer experience.
Innovation Solution
Utilize machine learning models trained on real-time and historical data to predict the likelihood of INF events at individual nodes, segmenting data to address class imbalance and overlap, and adjust inventory levels to minimize INF rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If real-time data processing is implemented to improve inventory prediction accuracy, then prediction accuracy is improved, but system complexity increases
Solution Approach 1:
The patent segments the inventory prediction system into multiple machine learning models, each trained on specific data segments (e.g., by item category, node type, or time period). This segmentation allows the system to process real-time data more efficiently by distributing computational load across specialized models, thereby improving prediction accuracy without overwhelming system complexity with a single monolithic model.
Solution Approach 2:
The system performs preliminary actions by pre-training machine learning models on historical data and pre-computing prediction parameters before real-time operation. This preliminary training phase allows the models to learn patterns from historical INF events, so that during real-time operation, the system can quickly apply these learned patterns to new data without requiring complex real-time computation, thus balancing accuracy with system simplicity.
2Measurement precision
If machine learning models are trained on segmented data to address class imbalance, then prediction accuracy is improved, but data processing time increases
Solution Approach 1:
The patent segments historical data into meaningful groups (e.g., by item category, node characteristics, or time periods) and trains separate machine learning models on each segment. This segmentation addresses class imbalance by ensuring each model is trained on data with relatively balanced classes for its specific context, improving overall prediction accuracy. The segmentation is performed once during the training phase, so it does not significantly impact real-time prediction processing time.
Solution Approach 2:
Different machine learning models are trained with different parameters and thresholds optimized for specific data segments or item categories. This local quality approach allows each model to be finely tuned to its specific context, improving prediction accuracy for each segment while maintaining efficient processing since each model operates independently on its specialized data type without requiring global reprocessing.
3Productivity
If suppression stock is reduced to optimize inventory allocation, then node utilization is improved, but INF rate increases
Solution Approach 1:
The system implements feedback by continuously monitoring actual INF events at fulfillment nodes and using this information to retrain and refine machine learning models. The models learn from real-world outcomes, adjusting their predictions to account for patterns that lead to INF events. This feedback loop allows the system to reduce suppression stock levels while maintaining low INF rates, as the models become increasingly accurate at predicting which nodes are likely to experience INF events without requiring excessive safety stock.
Solution Approach 2:
The system dynamically adjusts prediction parameters and thresholds based on changing conditions such as seasonal demand patterns, node performance variations, and item category characteristics. By changing these parameters adaptively, the system can optimize suppression stock levels for different time periods and contexts, improving node utilization during low-risk periods while maintaining reliability during high-risk periods without requiring uniformly high suppression stock across all scenarios.
Data Source
AI summary
Disclosed herein are systems and methods for predicting the probability of an inventory-not-found occurrence during fulfillment of an order, and for utilizing that predicted probability to improve order allocation applications and inventory systems. The disclosed systems and methods utilize machine learning predictive models to determine and score fulfillment performance of individual nodes by predicting their likelihood of an inventory-not-found event. At least some data received at the machine learning predictive application may be realtime data to reflect current conditions. In some examples, some data received at the machine learning application may be pre-computed data. The machine learning models generated by training at the machine learning application may include individual models used for different item categories. In some examples, a threshold suppression amount is utilized to adjust an amount of available inventory utilized to provide item availability (i.e. availability to place an order online) to a customer.