FinOps anomaly detection for near-immediate response

A major multinational pharmaceutical company

Challenge

With over 100 Azure Subscriptions FinOps team is not able to cope with cost anomalies in a timely manner: by the time when native Azure Cost Management tools identify cost anomaly and FinOps team make corrective action, unjustified spending for corresponding workloads may already last for days or even weeks.

Multiple responders are required for different cost anomaly origins, requiring additional reporting data for audit purposes and integration with other external systems.

Thus, the client needed a comprehensive technology partner to find a correct solution to the problems above.

Solution

To overcome limitations of native Azure Cost Management tools, our team optimized FinOps extraction method in the following way.

First, we developed custom Cost Anomaly Detection algorithm that detects cost anomalies in 24 hours frames - once outliers are identified by the algorithm, an upper fence calculation can be applied depending on cost owners’ goals. Then the team streamlined logs to include only essential data, improving performance and enabling detailed logging ensures all transactions are auditable.

GDC Services team developed scalable Logic App to support multiple endpoints across affiliates and integrate with external systems - each affiliate has dedicated workflows for billing, reporting, and compliance monitoring.

Also, custom workbooks with thresholds for alerts and workbook insights tailored for various cost centers were developed in order to provide insights into anomalies from the last 24 hours.

Finally, a 30-day trend graph was incorporated into the cost anomaly alerts, providing actionable insights.

Results

• Cost anomaly information became distributed to various affiliate admins the next day after anomaly occurs reducing financial impact from underutilized, unjustified and misconfigured Azure resources.

• FinOps extraction method for Azure was optimized for maximum efficiency in daily schedules, reducing detection curve by 75% - from 96h provided by native Azure Cost Management tools to 24h driven by our custom algorithm.

• The FinOps team repeatedly identified unjustified increases in SKU sizes across various Azure resources, as well as unused running compute capacity, resulting in cost savings of over $50,000 within 12 months after the solution was deployed.

• The company successfully saved over $6 000 in just two days after solution deployment, by promptly identifying unjustified and misconfigured Azure Data Factory pipeline runs, preventing significant accumulated costs.