Anomaly Detection System for eCommerce

Case Study

The e-commerce platform operations and analytics team of a prominent global brand wanted to develop an anomaly tracking system to monitor ongoing performance throughout their shopping and purchasing experience and detect issues to as soon as they surfaced. Leveraging our team’s expertise in designing and implementing machine learning solutions, we developed targeted anomaly alerting system with adjustable sensitivity at the metric level which runs detection hourly to surface issues as quickly as possible for revenue loss protection. The framework of the tool was designed to scale over time adding new metrics as needed and custom-tuning production metrics to handle state changes due to experience updates or changes at the data collection point.

Features of this system include:


KPI design and development to fit both business priorities and system requirements


Data analysis and visualization to identify anomaly thresholds and avoid under or over-alerting


Development and implementation of an hourly data collection pipeline to power anomaly alerting


Timeseries forecasting model development and optimization to identify anomalous values while continually learning and adapting for seasonality


Interactive investigation layer designed to allow stakeholders to quickly diagnose underlying issues


Technical project management and ongoing support to ensure alerting system continues to meet diverse requirements

End-to-end solutions that answer your most critical business questions.

Ready to put your data to work?