Regional Demand Estimation Using Bayesian Framework

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current systems for distributing items to sale and delivery locations struggle with handling large amounts of data, accommodating variations in demand, and making accurate predictions for new, seasonal, or slow-selling items, as they fail to consider the capabilities of item sale and delivery locations.

Innovation Solution

A system utilizing a Bayesian framework for demand estimation, which receives item identifiers, accesses overall demand forecast data, identifies geographical regions, evaluates demand share estimates, and generates demand estimations at various granularity levels using machine learning models to predict item utilization patterns.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If simple extrapolation techniques are used for demand distribution, then the system is easy to operate, but the prediction accuracy deteriorates for items with limited data

Engineering Contradiction:

Improvesystem operation simplicityVSAvoiddemand prediction accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The system changes the parameter of the prediction model from simple extrapolation to Bayesian framework, allowing dynamic adjustment of prediction accuracy based on data availability. For items with limited data, the system automatically adapts by using segment-level predictions and demand share estimates rather than direct item-level extrapolation.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The system introduces demand share estimates as an intermediary between overall demand forecasts and specific location predictions. This intermediary layer allows the system to handle items with limited data by using aggregated segment information to inform predictions at the item level, improving accuracy without complicating the overall system operation.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If the system considers regional variations and item capabilities for demand estimation, then the prediction accuracy improves, but the device complexity increases

Engineering Contradiction:

Improvedemand prediction accuracyVSAvoidsystem structure complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments items into different categories (new items, seasonal items, slow-selling items, fast-selling items) and applies appropriate prediction methods to each segment. This segmentation allows the system to handle complexity by treating different item types differently, improving prediction accuracy for each category while managing overall system complexity through modular processing.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system dynamically adjusts the level of detail in predictions based on data availability and item characteristics. For items with sufficient data, direct item-level predictions are made; for items with limited data, the system dynamically shifts to using segment-level predictions with demand share adjustments, thereby adapting complexity to need rather than applying a fixed complex model to all items.

Inventive Principle:

Principle #15Dynamics

3Measurement precision

If the system uses Bayesian framework with multiple data sources, then the demand estimation accuracy improves, but the loss of time for processing increases

Engineering Contradiction:

Improvedemand estimation accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by pre-calculating and storing segment-level demand estimates and demand share data. When making predictions for specific items, the system leverages these pre-computed values rather than calculating everything from scratch, significantly reducing processing time while maintaining the benefits of the comprehensive Bayesian framework.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system applies partial action by selectively using different levels of prediction detail based on data availability. For items with sufficient data, only item-specific calculations are performed; for items with limited data, the system uses pre-computed segment-level predictions with minimal additional processing to adjust for demand share, thereby reducing overall processing time while maintaining accuracy where possible.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20230128417A1Systems and methods for regional demand estimation

Publication Date: 2023.04.27 COUPANG CORP

AI summary

Systems and methods, and computer readable media for inventory demand estimation of a region are disclosed. The method receives an item identifier associated with an item for demand estimation. The method may then access overall demand forecast data for the item and identify geographical regions, and evaluate a demand share estimate of the item in the geographical regions. The method may also determine a set of item identifiers associated with a segment of items related to the item and determine the demand estimation for the segment of items. The method may then generate demand estimation of the item at the geographical regions using a Bayesian framework with demand share of the item and demand estimation of the segment of items in a geographical region, and overall demand forecast data for the item as input. The method may use the generated demand estimation at a region to generate demand estimation in other encompassing geographical regions.