How do you analyze cloud cost and FinOps data in Metabase?
Cloud cost tools hold the financial record of your infrastructure: every charge, commitment, tag, and anomaly. In Metabase, land that record in a SQL warehouse — billing exports make this unusually easy — and model layers forcost rollups, allocations, commitments, and unit costs. Those layers support dashboards for spend, team allocation, commitment coverage, anomalies, and cost per customer — next to the revenue and product data the rest of the company already reports on.
TL;DR — Billing data belongs in a warehouse, and the clouds will put it there for you: AWS via CUR 2.0 exports to S3 + Athena, Google Cloud via the BigQuery billing export. Model amortized vs. unblended cost once, keep credits separate, and treat the untagged-spend rate as the ceiling on every allocation view.
Which tools does this cover?
- AWS Billing — billing line items, daily cost rollups, Savings Plans and Reserved Instance commitments, budgets
- Google Cloud Billing — billing export cost lines, projects and services, labels, credits and committed use discounts
- Vantage — cost report rows, cost reports, providers and integrations, virtual tags
- Kubecost — allocation rollups, namespaces and workloads, asset costs (nodes, disks, network), idle and unallocated cost
- CloudZero — dimension cost slices, dimensions (customers, products, features), unit cost metrics, anomaly events
For the infrastructure side of the same story, see Kubernetes, Amazon EKS, and GKE — utilization and capacity data joins cost data on cluster, namespace, and workload keys.
What is the shared cloud cost data model?
| Concept | Common terms | Used for |
|---|---|---|
| Billing line item | CUR line item, export row, usage record | The atomic source of truth; drill-downs |
| Cost rollup | Daily cost by service, account, team | Spend trends, growth, budget burn |
| Allocation | Tag, label, dimension, cost category | Team and product cost ownership |
| Commitment | Savings Plan, RI, CUD | Coverage, utilization, effective savings |
| Unit cost | Cost per customer, feature, request | Unit economics and margins |
| Anomaly event | Spike, deviation, alert | Cost incident response |
| Budget | Budget, forecast, plan | Burn rate and variance |
| K8s allocation | Namespace, workload, idle cost | Cluster cost and efficiency |
How do you connect cloud cost tools to Metabase?
- Billing export route — the durable anchor: CUR 2.0 to S3 + Athena for AWS, the BigQuery export for Google Cloud. The cloud maintains the pipeline; you model on top.
- MCP + CLI route — pull a scoped, summarized cost slice through a vendor MCP server, save CSV, and load it with
mb upload csvfor fast exploration. - Platform sync route — sync normalized costs, virtual tags, and recommendations from FinOps platforms via connector or API, and reconcile against raw exports quarterly.
What can you analyze across these tools?
- Total cloud spend— amortized spend per period, with credits modeled separately
- Cost growth rate — month-over-month change, attributed to services and teams
- Commitment coverage rate — covered share of eligible spend, read with utilization
- Effective savings rate — realized savings against on-demand-equivalent cost
- Untagged spend rate— the ceiling on every allocation dashboard
- Budget burn rate — budget consumed vs. period elapsed, per team
- Cost per service — provider services and application services, mapped deliberately
- Cost per customer— allocated cost joined to revenue for margin per customer
Which dashboards should you build?
- Cloud spend overview — total spend, service and account breakdowns, growth, and budget burn
- Cost allocation by team and tag — ownership views with the untagged bucket visible
- Commitment coverage — Savings Plans, RIs, and CUDs: coverage, utilization, expirations
- Cost anomalies — spikes, deviations, and response times
- Kubernetes cost — namespace and workload cost with idle share
- Cost per customer — unit economics with revenue joined
Legacy KPI-dashboard guides cover adjacent ground: AWS monitoring, multi-cloud, storage costs, and cost performance indicators.
Common mistakes
Mixing amortized, unblended, and net-of-credits cost.→ Pick a convention per view and label it — trends on one basis and totals on another never reconcile.
Pointing dashboards at raw CUR line items.→ Billion-row tables make slow, expensive cards. Build daily rollup models and reserve line items for drill-downs.
Reporting allocation while ignoring untagged spend.→ Show the untagged rate on the same dashboard — allocation percentages are only as true as tag coverage.
Celebrating commitment coverage without utilization.→ Over-commitment looks like success in a coverage-only view. Track unused commitment dollars explicitly.
Building unit economics before the cost side is trustworthy.→ Cost per customer inherits every allocation error underneath it. Get rollups, tags, and shared-cost rules right first.
Related
Analytics
Integrations
Data sources
FAQ
What is cloud cost analytics?
Cloud cost analytics is reporting built on your billing data — spend by service and team, commitment coverage, anomalies, and unit economics — modeled in a SQL warehouse and analyzed in Metabase. FinOps platforms answer "how do we optimize rates?"; cost analytics in your own BI answers "what does our product cost to run, per team and per customer, next to revenue?" See the cloud cost analytics overview for the full build.
Does Metabase connect natively to AWS or Google Cloud billing?
Metabase reads SQL databases and warehouses — and that's exactly where cloud billing data wants to live. AWS delivers CUR 2.0 exports to S3, queried through the Athena driver; Google Cloud writes its billing export straight into BigQuery, which Metabase queries directly. No middleware is involved in either path.
Do I need a FinOps platform, or are raw billing exports enough?
Raw exports (CUR 2.0, BigQuery export) plus warehouse models cover spend, allocation, commitments, and anomalies for a single cloud — that's the build most teams should start with. Platforms earn their keep on top: Vantage for multi-cloud normalization and virtual tags, CloudZero for cost-per-customer dimensions, Kubecost for inside-the-cluster allocation.
What is the difference between cost visibility and unit economics?
Cost visibility answers "what are we spending?" — spend by service, account, and team. Unit economics answers "what does a unit of business cost?" — cost per customer, per feature, per request — and needs cost data joined to customers, usage, and revenue. The join is where a BI tool beats a cost console: Metabase puts billing data next to the business tables that give it a denominator.
Which tools does the cloud cost category cover?
Five guides: the two cloud billing anchors (AWS Billing via CUR 2.0 and Athena, Google Cloud Billing via the BigQuery export), Vantage for multi-cloud FinOps, Kubecost for Kubernetes allocation, and CloudZero for unit economics. All map onto a shared model — daily cost rollups, allocations, commitments, and unit costs.
Which dashboards should you build first from billing data?
Start with the cloud spend overview — it needs only daily rollups from a billing export. Add team and tag allocation once tag hygiene supports it, commitment coverage when Savings Plans or CUDs are in play, then anomalies, Kubernetes cost, and cost per customer as the practice matures.