Location Recommender Engine
A startup in the social events space came to PressW with a unique challenge. This company wanted to build a recommendation engine that would deliver highly relevant location suggestions for users while on their application. The locations needed to be contextually relevant to the social event, factor in each user’s individual preferences, and needed to be near real-time in order to drive their core KPI, number of events made on their platform.
We built our client a custom recommendation engine that aggregated locations from a number of POI data providers and delivered real-time suggestions during a user’s plan creation. We built a two step system that first matched recommendations based on the event details, distance, and reviews before then passing through a user preference model that further increased the relevancy and improved rankings of the relevant locations. To deal with the real-time requirement, we built this system utilizing cosine similarity vector search hosted through server-less functions that were able to handle a load into the millions of concurrent requests.
Our solution dramatically modified our client’s location relevancy and ultimately resulted in a massive increase to their core KPI. Product surveys they conducted showed their user’s felt a significant increase in suggestion relevancy and helped build trust with the product and platform.
Our client saw a 70% increase in location relevancy from their former system, and a 42% increase in plans created using suggested locations.