Manuel Porras

Ad Campaign Anomaly Detection

Protecting ad spend across 30+ client campaigns

The Problem

A marketing agency managing PPC campaigns for 30+ clients had no early warning system for budget issues. Overspend was discovered days after it happened. Lead count drops went unnoticed until client calls. The team was checking dashboards manually across multiple platforms, and anomalies slipped through the cracks.

The Approach

I built a real-time monitoring pipeline that collects spend and performance data daily from Google Ads, computes rolling baselines, and flags statistical deviations. The system uses adaptive thresholds that account for day-of-week patterns and seasonal trends.

Anomalies are classified into severity tiers (critical, warning, info) based on deviation magnitude and business impact. Critical alerts escalate automatically via Slack with contextual data: what changed, by how much, and what the baseline was. The system also tracks month-to-date budget pacing with projected end-of-month spend calculations.

A unified dashboard shows all clients with traffic-light status indicators, budget progress bars, and drill-down views per platform. Inactive clients are automatically excluded from alerting based on a Google Sheet sync that reads the team’s budget spreadsheet.

The Outcome
30+
Campaigns monitored daily
3-tier
Anomaly severity classification
Auto
Slack alerts with context
Tech Stack
Node.jsPostgreSQLGoogle Ads APISlack APIGoogle Sheets APIExpress
Related

← Back to all projects