Yelp transitioned thei r main web application from a single monolith to a service-oriented archite cture in order to scale application development. This transition distribute d data-producing services across the organization and introduced new analyt ics challenges. Data silos emerged as a consequence of having disparate dat a sources that are difficult to join. Each data silo has an incomplete view of the full product which limits feature developers from understanding the effects of their changes on the product as a whole.
\nStreamin g data infrastructure has allowed the data side of Yelp to scale with the a pplication side. Standardizing a data logging and connector ecosystem has m ade it easy to stream data between data stores. This allows for rapid itera tion through A/B testing by surfacing experiment-specific and company-wide metrics, enabling experimenters to safely deliver the most impact.
\nThis makes it easier and faster for developers to unlock and process rich d ata streams, improve the analytics story, and provide greater vis ibility into a product's health.
\nWe will cover:
\n\n The building blocks of a streaming ecosystem\n The network effects unlocked by event logging at scale\n How streams enable safe and quick experimentation iteration cycles\n\n DTSTART:20181011T105000 SUMMARY:A Stream-First Approach to Analytics and Experimentation DTEND:20181011T114959 LOCATION: See Description END:VEVENT END:VCALENDAR