Blog

It’s a familiar story for many of us working with marketing teams: the endless cycle of exporting customer lists, manually re-formatting them for Google Ads, then for Meta Ads, then for Twitter, and so on. This process is not only tedious but also error-prone and a significant drain on valuable development time.

For one of my clients, I decided to build a streamlined solution to this exact problem. I’ll walk you through the audience engine, a simple, centralized API that connects my client’s BigQuery data warehouse directly to their ad platforms, automating audience updates with a single request. 🚀📈

Automated anomaly detection is a critical component for any production system. Whether you’re tracking user engagement, ad spend, or infrastructure load, the ability to catch unexpected spikes or dips isn’t just a convenience—it's essential for maintaining a healthy service.

This article details a recent client project focused on migrating their anomaly detection system from a managed, “black-box” service, the GCP Timeseries Insights API, to a more transparent and robust solution within their BigQuery data warehouse.

I’ll walk through the initial appeal of the simple API and the real-world challenges that ultimately prompted the switch. More importantly, I will provide a practical, step-by-step guide on how we implemented this new, more controllable anomaly detection system for the client using BigQuery ML's powerful ARIMA_PLUS modeling capabilities. 📈