Anomaly Detection for Post-Deployment Defects

Accelerating the rollout of new features in a technical environment can expose issues. To mitigate these challenges, our client relied on Oxford’s support. 

INDUSTRY
Vacation Rental Services

SERVICES
Machine Learning
Automated Pipelines
Containers and Orchestration
Cloud Enablement

SKILLS
Data Scientist
AI Engineer
Cloud Engineer
Automation Engineer

The Challenge
Our client, a world leader in vacation rental services, wanted to accelerate the rollout of new features in their technical environment. Accelerating the deployment exposed many issues with our client’s transaction pipeline that were previously undetected by testing. The client wanted to find a way to discover and address these post-deployment issues that only impacted hard to detect segments of potential website users, and that may have been largely undiscoverable through manual analysis.

The Solution
Our team applied their expertise in both big data and machine learning to help our client create an anomaly detection service, reusable data science container and orchestration solution. The anomaly detection service was designed to find anomalies in the data with both elasticity and scalability to quickly and efficiently detect anomalous patterns or behaviors, along with the affected user segment.

The Result
The anomaly detection service provided the client with a safety net that allowed them to release new products, services, and features as quickly as possible with the best possible quality and quick discovery of post-deployment defects. The service also provided the ability to look for these issues at the macro level or zoom in on micro groups such as device type, browser, geography or audience segment to help identify the root cause of the problem. Additionally, the service was created in an open manner so that anomaly detection can be used for any time of time-series data pipelines.

Share This