ML Item Weight Prediction for Accurate Picker Compensation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Online concierge systems face inaccuracies in item weight classification due to relying on retailer-provided weights, leading to under- or over-compensation of pickers, as they lack physical possession of items and cannot verify the accuracy of weight information.

Innovation Solution

The system trains a weight prediction model using item attributes from retailer catalogs to predict item weights accurately, classifying items as heavy or light based on a threshold, and updates the catalog with refined weights through manual verification when discrepancies are detected.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If the online concierge system uses retailer-provided weights to determine item classification, then the system can process orders without physical possession of items, but the weight accuracy deteriorates leading to under- or over-compensation of pickers

Engineering Contradiction:

Improveorder processing capabilityVSAvoiditem weight accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent introduces an intermediary machine learning model that acts as a mediator between retailer-provided weights and compensation decisions. The model takes multiple item attributes (keywords, brand, manufacturer, type, price, quantity, size) as input and predicts accurate weights without requiring physical possession of items. This intermediary system resolves the contradiction by providing weight accuracy comparable to physical measurement while maintaining the operational efficiency of remote order processing

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the mechanical weighing system (physical possession and measurement of items) with a computational system using machine learning. Instead of physically weighing items to determine accuracy, the system uses trained models that process item attributes and predict weights with high accuracy, substituting mechanical measurement with intelligent algorithms

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Device complexity

If the system classifies items based on inaccurate retailer weights, then compensation processing is simplified, but picker morale deteriorates due to under-compensation

Engineering Contradiction:

Improvecompensation calculation complexityVSAvoidpicker morale

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent replaces simple but unreliable threshold comparison based on retailer weights with a machine learning-based weight prediction system. The model processes multiple item attributes and outputs predicted weights that are then used in threshold comparisons, maintaining the simplicity of the compensation calculation process while dramatically improving accuracy and picker morale through more reliable weight estimates

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Ease of manufacture

If the system uses catalog weights from retailers, then data collection is simplified, but cost increases due to over-compensation for inaccurate heavy item classification

Engineering Contradiction:

Improvedata collection processVSAvoidoperational cost

Core Design Contradiction:

Ease of manufactureVSLoss of energy

Solution Approach 1:

The patent introduces a machine learning model as an intermediary layer between retailer catalog data and compensation calculations. The model processes item attributes from the catalog and predicts accurate weights, allowing the system to maintain simple data collection from retailers while avoiding the costs of over-compensation. The intermediary model filters and refines the catalog data to eliminate systematic errors that lead to over-payment

Inventive Principle:

Principle #24Intermediary (Mediator)

4Adaptability or versatility

If the system relies on retailer-provided weights, then physical inventory management is avoided, but measurement accuracy deteriorates

Engineering Contradiction:

Improvesystem flexibilityVSAvoidweight verification capability

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent replaces physical inventory management and direct weight measurement with a virtual system using machine learning models. The system maintains flexibility by processing digital catalog data without physical item handling, while simultaneously achieving measurement precision through trained models that predict weights based on multiple item attributes, effectively substituting physical verification with intelligent computation

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS12572911B2Predicting item weights using a trained machine learning model

Publication Date: 2026.03.10 MAPLEBEAR INC

AI summary

An online concierge system compensates pickers who fulfill orders including one or more items based in part on weights of the items included in an order. Because the online concierge system does not physically possess the items that are obtained, the online concierge system cannot directly weigh the items and weights specified for items in a catalog from a retailer may be inaccurate. To more accurately determine weights of items, the online concierge system trains a weight prediction model to estimate an item's weight from attributes of the item and uses the output of the weight prediction model to determine compensation to a picker. The weight prediction model may output a predicted weight of an item or a classification of the item as heavy or light. Where discrepancies are found between a predicted weight and the catalog weight of an item, additional information about the item is obtained.