Designing a Dynamic User Segmentation and Automation System for Growth Operations

The author, a growth operations engineer, shares a story about an A/B test where coupons were sent to users who hadn't ordered for over a week, highlighting the need for precise user targeting and cost‑effective promotion.

From this scenario, two core problems are identified: (1) finding the right users quickly and integrating them with business systems, and (2) making routine operational rules reusable and automatable.

To solve the first problem, the concept of a "user group" is introduced. A dynamic query system is built with three components: data routing (providing field locations and types), a query generator (translating UI configurations into SQL or Elasticsearch queries), and group reuse (static and dynamic persistence of user ID sets).

The second problem is addressed by decoupling user selection from rule execution. Once a group is defined, operators can attach any delivery channel (push, SMS, coupon) and schedule the rule with a cron‑like scheduler, achieving fully automated daily activation campaigns.

The article then discusses data and application extensions, noting that the solution puts pressure on OLAP workloads and requires a separate read‑only replica. Real‑time needs are met by adopting a Lambda architecture that combines batch‑processed historical views with a speed layer updated via Kafka, Spark Streaming, and Canal.

With near‑real‑time data, the grouping system can also drive passive features such as dynamic ad placement, automatically adjusting content based on user behavior without code changes.

Finally, the author acknowledges that the current implementation covers only part of the problem space and points to future work on feedback normalization and automatic rule optimization to achieve a complete closed‑loop system.