Building an Audience Builder: A Marketer Automation System Created by an Analyst

Previously, setting up marketing campaign audiences required complex communication and manual work among marketers, data personnel, and development teams. This process resulted in a persistent bottleneck where target lists failed to stay up-to-date, reducing campaign efficiency.

To address this, Audience Builder was introduced. When marketers set conditions via natural language filters or the UI, the system converts them into SQL, executes them on BigQuery, and automatically sends the results to marketing tools. This reduced inter-organizational wait times and enabled timely campaign operations.

Despite having no development experience, the analyst directly implemented frontend and backend logic using AI agents and Vibe Coding. AI was utilized for brainstorming, spec reviews, and repetitive QA tasks to maximize development productivity.

The data structure was dualized into PostgreSQL for metadata management and BigQuery for processing large-scale audience lists. It dynamically builds SQL at runtime by referencing the user feature store catalog and processes extraction tasks of tens of millions of rows via asynchronous workers.

Custom pipelines were built using S3, Kafka, and REST API to handle different data delivery methods for each marketing tool. Notably, the design ensures the Baseline version is updated only upon successful delivery, preventing data loss risks.