Multi-Objective Ranking Algorithm for Booking Diversity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online computer systems for experience reservations prioritize maximizing total bookings without considering business needs such as inventory balancing and social goals like diversity and inclusiveness, leading to suboptimal results for hosts and users.
Innovation Solution
A computer-implemented method that generates a ranked list of search results by incorporating multiple objectives, including inventory balancing and diversity, through a multi-objective function that decomposes serving vectors to optimize beyond booking metrics, ensuring listings are presented in a way that balances occupancy rates and offers diverse experiences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the system ranks listings according to a single objective of maximizing total expected bookings, then the total number of bookings is improved, but inventory balancing and diversity goals deteriorate
Solution Approach 1:
The patent segments the single booking objective into multiple separate objective functions: (1) expected number of bookings, (2) inventory balancing metric, and (3) diversity metric. Each objective is calculated and weighted independently, then combined to form a comprehensive ranking score. This segmentation allows the system to optimize for bookings while simultaneously considering inventory and diversity goals.
Solution Approach 2:
The ranking system is transformed into a multi-functional system that simultaneously performs multiple objectives: maximizing bookings, balancing inventory across hosts, and ensuring diversity in search results. The multi-objective function serves as a universal framework that integrates these different goals into a single ranking mechanism.
2Productivity
If the system prioritizes maximizing total bookings, then booking volume is improved, but host revenue and opportunity cost deteriorate due to low occupancy rates
Solution Approach 1:
The system incorporates feedback from historical booking data and occupancy rates into the ranking process. The inventory balancing metric uses feedback about each host's current occupancy rate to adjust rankings, boosting listings from hosts with lower occupancy to ensure they meet minimum thresholds. This feedback mechanism allows the system to dynamically adjust rankings to protect host revenue while maintaining overall booking volume.
3Productivity
If the system displays multiple experiences of the same category consecutively to maximize bookings, then booking conversion is improved, but user experience and diversity deteriorate
Solution Approach 1:
The patent applies local quality by making the ranking criteria vary depending on the specific listing and context. The diversity metric evaluates each listing's contribution to overall category diversity, and the ranking adjusts locally for each position in the search results. This allows the system to maintain high booking conversion while strategically introducing diverse categories at different positions in the results.
Solution Approach 2:
The system changes the ranking parameters by incorporating category diversity as a explicit factor in the multi-objective function. Instead of using a single parameter (booking probability), the system uses multiple parameters including expected bookings, inventory balance, and diversity metrics. This parameter change enables the system to optimize for both conversion and diversity simultaneously.
Data Source
AI summary
A computer implemented method for incorporating multiple objectives in a ranked list of search results includes receiving a search query from a client device, accessing a set of stored listings for goods or services and probabilities of serving the listings, defining a serving vector as a probability distribution over the set of listings, providing a serving vector as input to a multi-objective function, decomposing the multi-objective function into one or more objective functions, generating a ranked list of the listings based at least in part on the serving vector that maximizes the decomposed multi-objective function, and providing the listings to the client device according to the order of the ranked list. Each objective function addresses a different goal in an overall diversity optimization.