Anomalies

North's hands free cost & usage anomaly detection system. Running for you with no human setup needed.

Anomaly Detection

North’s Anomaly Detection is designed for maximum precision, leveraging resource-level and hour-level cost and usage data. This means we’re able to surface extremely granular and accurate anomalies—letting you catch issues that would be invisible to conventional monitoring.

How It Works

We begin by aggregating and analyzing your cloud data at the usage type level—a specific operation or group of resources for a specific service and provider. Whenever our models flag an anomaly at this usage type level, we break it down further: if the anomaly can be linked to specific resources, North will display those resource-level anomalies as well, allowing you to pinpoint exactly which resources are causing the exception. Note: Not every usage anomaly is traceable to an individual resource, but every anomaly is visible at the usage-type resolution for full auditability.

Smart Pattern Recognition

North doesn’t just surface every spike or dip: our advanced pattern recognition models look at trends and the frequency of repetitions to avoid noise from normal patterns—like new resources being spun up, or expected spikes during specific times of the month. Our system is engineered to avoid unnecessary alerting so you only see true, actionable anomalies.

When digging into any anomaly, users can explore deep context and history. Every anomaly can be shared instantly with teammates via Slack, email, or Jira, making it easy to coordinate investigation and resolution.

Learning & Continuous Improvement

Customer feedback shapes our anomaly detection. When issues are resolved, masked/ignored, or marked as not relevant, that data helps retrain and tune the ML models for your specific environment—so over time your alerts become more accurate and tailored to your organization.

You can further refine North’s anomaly reporting through the settings:

• Minimum Dollar Threshold: Only be alerted to anomalies that exceed your defined minimum cost impact.

• Max Alert Volume: Limit how many anomalies are surfaced per period, so you can dial in noise vs. visibility to match your team’s capacity.

Key Features

• Resource-Level Anomaly Detection: Hourly and daily analysis pinpoints even subtle outliers, right down to the resource

• Usage-Type Grouping: All anomalies are tracked by usage-operation for complete context

• Pattern Recognition: Intelligently suppresses alerts for routine or expected patterns (e.g., new deployments, known spend cycles)

• Precision Alerting: Only actionable items surface—reducing noise

Anomaly Types

Cost Spikes

• Detected when spend on a usage type or resource jumps beyond statistically expected levels, using z-score and multi-dimensional time series analysis.

Variations & Sudden Changes

• Flags large step changes in spend for specific usage types or resources, taking into account normal day-to-day variability.

Dips/Drop-offs

• Surfaces sudden decreases that may point to resource deletions, lapsed reservations, or accidental shutdowns.

Integration & Collaboration

• Slack, Email, and Jira Integration: Share, escalate, and collaborate on any anomaly with a single click.

• Dashboard: Deep-dive, interactive exploration of each anomaly, all the way from usage group to individual resource.

Customization & Controls

• Configurable Thresholds: Adjust minimum dollar alert, sensitivity, and alert volume in preferences

• Masking and Reason Tracking: Mask anomalies you know are safe, add masking reasons, and maintain an audit trail for compliance and training

Best Practices

• Review Regularly: Monitor the anomaly dashboard to quickly catch new issues

• Refine Sensitivity: Tune settings to match your team’s response bandwidth

• Collaborate on Resolution: Use deep context, sharing, and masking features to coordinate effective follow-up

• Link to Events: Correlate anomalies with product launches, deploys, seasonal business cycles, or infrastructure projects