Surfacing time-sensitive stock market intelligence

AfterHour is a dynamic platform designed to provide social intelligence and timely market insights. With features like real-time stock recommendations, user-generated content, and social interaction, AfterHour aims to create a vibrant community of traders and investors. The platform hosts an exponentially growing number of new posts daily.

The stock market moves fast. That means an app focussed on providing trading social intelligence and alpha through a ‘for you’ feed needs to operate in real-time. AfterHour’s goal was to be the primary source of timely trading information users couldn’t find anywhere else, surfacing unique recommendations based on a users’ holdings as soon as they opened the app. When creators posted recommendations of a trade, AfterHour needed that information to be disseminated to the right interested users in an instant.

AfterHour wanted a configurable solution where they could weight content according to recency, position and stock specifics, user behavior, and other factors, while maintaining real time latency. They needed a solution that could scale with their business, while avoiding complex data engineering.

Real-time re-ranking is what allows apps like TikTok or IG Reels to have relevant content ranked while you're interacting with the application. For example, if you're a new user and you see a cat video and like it, for the rest of that session you're more likely to see more cat videos. Real-time re-ranking is what gives products a reactive, dynamic feeling, and is critical to build the most relevant experience for end-users.

Afterhour partnered with Shaped to help re-rank their 'for you' feed and build a reactive, and highly relevant experience. 

How does Shaped do it? The real-time in real-time ranking is actually referring to several components. It's the real-time ingestion of interactions (i.e. the feedback), the real-time feature creation from these interactions (i.e. the transforms) and finally the real-time re-ranking of results (i.e. the model inference).

Below we outline these components in further detail:

1. Real-Time Ingestion Real-time ingestion involves the immediate intake and processing of data as it is generated. One of the most powerful features of Shaped is its real-time data connectors, which allowed AfterHour to easily connect their data-streaming and Customer Data Platform sources and stream data into Shaped in real-time. This low-latency process ensures data is quickly available for accurate and timely stock recommendations and updates.

2. Real-Time Reranking Real-time reranking updates the order of recommendations in AfterHour's feed based on the latest data and user interactions. Shaped’s machine learning models assess the relevance of each recommendation in real-time, considering factors like recency, user behavior, and stock specifics. This approach ensures users receive the most relevant and timely trading information.

3. Real-Time Stateful Window Transformations Real-time isn’t just about reranking a specific user’s session, but also about aggregating all user interactions over a time period to understand trends. Real-time stateful window transformations for AfterHour involve processing user interaction data streams while maintaining context over a specific time window e.g the last five minutes. Shaped’s proprietary framework performs complex, low-latency computations on this windowed data, allowing the system to quickly analyze recent trends and user behavior. An example for AfterHour is a feature ‘num comments in the last minute’. A high number may represent an interesting post, such as a news event, and the post will be ranked higher accordingly.

With these three capabilities optimized for real-time, AfterHour's feed can re-rank content dynamically for each user based on the latest user interactions and market trends within seconds, easily keeping up with the speed of the market.

The implementation of real-time recommendations for AfterHour significantly enhanced user experience by delivering timely and relevant stock suggestions. By leveraging Shaped's advanced machine learning models and real-time processing capabilities, AfterHour ensured that users received valuable recommendations based on the freshest data and current market trends. This immediate responsiveness not only kept users engaged but also empowered them to make informed trading decisions quickly. The result was a dynamic, responsive and socially vibrant platform that met the demands of a fast-paced financial market, ultimately leading to increased user satisfaction, retention and potentially market gains.