Insights

CostPilot Insights are automatically generated recommendations that identify specific optimisation opportunities in your cluster. Each insight includes the problem identified, estimated monthly savings, and recommended actions.

How insights are generated

Insights are generated automatically — you don’t need to trigger them manually. CostPilot’s processing pipeline analyses your metric data and generates a new set of insights whenever sufficient data has been collected (approximately every 5,000 metric records, which for a typical cluster means every few hours).

The insight engine runs multiple analysers in parallel across six categories: Cost, Efficiency, Resource, Optimisation, Capacity, and Reliability. Results for all analysers are produced in roughly 200ms regardless of how many run concurrently.

Insight categories

Cost

Identifies workloads and namespaces generating disproportionate spend relative to their actual usage, including idle cost patterns and allocation inefficiencies.

Efficiency

Flags workloads where actual CPU or memory usage is consistently below their resource requests. Each finding includes the workload name, current vs. used resources, and the estimated monthly savings from right-sizing.

Resource

Examines raw resource consumption patterns — CPU throttling, memory pressure, BestEffort pod risks, and missing resource limits.

Optimisation

Cloud-provider and instance-level recommendations: spot adoption opportunities, cheaper equivalent instance types, and commitment discount candidates.

Capacity

Node-level analysis: underutilised nodes, consolidation opportunities, node pool sizing, and autoscaler configuration gaps.

Reliability

Identifies high-availability risks: single-replica Deployments in production namespaces, workloads without resource limits, and namespaces approaching quota limits.

Tier coverage

The analysers that run depend on your plan tier:

Analyser	Basic	Pro / Max
Idle Cost	✓	✓
Basic Efficiency	✓	✓
Node Health	✓	✓
Reliability	✓	✓
Advanced Efficiency	—	✓
Attribution	—	✓
Cloud Optimisation	—	✓
Resource Distribution	—	✓
Namespace Patterns	—	✓
Quick Wins	—	✓
Temporal Trends	—	✓

Severity levels

Each insight is classified by urgency:

Severity	Colour	Meaning
Critical	Red	Significant ongoing waste or risk. Act immediately.
Warning	Amber	Meaningful optimisation opportunity. Schedule for this sprint.
Info	Teal	Minor improvement or informational finding. Review when convenient.

Insight statuses

Status	Meaning
Active	New or not yet reviewed
Acknowledged	Seen and noted — will action later
Dismissed	Not relevant — won’t action
Resolved	Action taken — savings realised

The Insights page is split into tabs matching these statuses. Use Active as your default working view. Move insights to Acknowledged to clear them from the active list without losing them. Use Dismissed for insights that are genuinely not applicable to your situation (e.g. a “single replica” warning for a deliberately singleton service).

Acting on insights

Each insight card shows:

• Title and description — What was found and why it matters
• Affected resource — The cluster, namespace, or workload involved
• Potential savings — Estimated monthly savings in your configured display currency
• Recommended action — Specific steps to resolve the issue

The Insights overview page groups findings by category, showing a savings opportunities panel and a “requires attention” panel for critical or warning-severity items.

Relationship to Signals

When the Analysis page detects a cost anomaly in the Signals tab, a corresponding insight is often generated to explain and quantify it. If an insight references an unusual pattern, the underlying signal is visible in Analysis → Signals.