Merchant Affinity Profiling Using Transaction-Based ML Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Customer surveys for analyzing shopping trends are time-consuming, expensive, and prone to sampling errors, making them inefficient for inventory management and marketing efforts.
Innovation Solution
A system utilizing machine-learning models to generate affinity profiles based on transaction data, without requiring personal information, by training models to create user and merchant embeddings, determining similarities, and updating them based on purchase history to predict user behavior and merchant interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If customer surveys are used to analyze shopping trends, then shopping trends can be analyzed, but the process becomes time-consuming and expensive
Solution Approach 1:
The patent replaces the mechanical survey system with a machine learning-based automated analysis system. Instead of manually conducting surveys to gather shopping trend data, the system uses trained machine learning models that process transaction data automatically to generate affinity profiles and predict shopping behaviors, eliminating the time-consuming survey process while maintaining or improving analysis accuracy
Solution Approach 2:
The patent creates synthetic affinity profiles and user journey representations that copy and generalize from historical transaction patterns. By training models on past data and generating representative profiles, the system can analyze shopping trends without needing to conduct new surveys, effectively copying insights from historical data rather than gathering new data through time-consuming surveys
2Measurement precision
If customer surveys are used to analyze shopping trends, then shopping trends can be analyzed, but the cost increases
Solution Approach 1:
The patent replaces expensive survey methodologies with automated machine learning models that process existing transaction data. The system eliminates the need for costly survey design, distribution, and analysis by using trained models to automatically generate affinity profiles and shopping trend insights from readily available transaction records
Solution Approach 2:
The system enables self-service analysis by automatically generating affinity profiles and shopping trend insights without requiring external survey intervention. The machine learning models autonomously process transaction data, generate user journey predictions, and provide actionable insights, eliminating the need for paid survey services
3Measurement precision
If customer surveys are used to analyze shopping trends, then shopping trends can be analyzed, but sampling errors occur
Solution Approach 1:
The patent creates comprehensive affinity profiles that copy and represent entire user populations based on transaction data patterns. By generating profiles from actual transaction records rather than survey samples, the system captures authentic shopping behaviors without the sampling biases inherent in survey methodologies
Solution Approach 2:
The system incorporates feedback loops where model predictions are continuously refined based on actual transaction outcomes. The machine learning models learn from the difference between predicted and actual shopping behaviors, continuously improving accuracy and eliminating sampling errors through iterative refinement on comprehensive transaction data
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system comprising one or more processors and a computer-readable, non-transitory medium including instructions which, when executed by the one or more processors, cause at least one of the one or more processors to obtain transaction data including a first merchant and a first user, and train a machine-learning model using the transaction data by executing the machine-learning model using as input the transaction data to generate a first merchant embedding corresponding to the first merchant and a first user embedding corresponding to the first user determining a similarity between the first merchant embedding and the first user embedding, and updating the machine-learning model based on whether the first user made a purchase at the first merchant within a predetermined time period.