Metric

What is untagged spend rate?

Definition

Untagged spend rate is the share of cloud cost carrying no owner tag — no team, project, or cost-center attribution. It is the ceiling on every allocation dashboard: if 30% of spend is untagged, team cost views are at best 70% true. Mature FinOps practices treat it as a first-class KPI with an owner.

Formula: Untagged spend rate = cost with no owner tag / total cost × 100

What data do you need?

  • Billing line items with tags and labels preserved through the pipeline
  • The list of required tag keys (team, environment, service)
  • A tag-to-owner mapping that normalizes variants and typos
  • Resource identifiers so top untagged resources can be assigned
  • Untaggable-spend classification (some charges can't carry tags)

SQL pattern

Untagged spend rate and the biggest untagged resourcesPostgreSQL
SELECT
  date_trunc('month', usage_date) AS month,
  ROUND(
    100.0 * SUM(amortized_cost_usd) FILTER (WHERE team_tag IS NULL)
    / NULLIF(SUM(amortized_cost_usd), 0), 1
  ) AS untagged_spend_pct,
  ROUND(SUM(amortized_cost_usd) FILTER (WHERE team_tag IS NULL), 2)
    AS untagged_usd
FROM billing_line_items
WHERE line_item_type NOT IN ('Credit', 'Refund', 'Tax')
GROUP BY 1
ORDER BY 1;

Common pitfalls

Counting untaggable charges against the rate.→ Support fees, taxes, and some shared services can't carry tags — classify them separately so the rate measures fixable gaps.
Treating tag variants as different teams.→ Normalize 'Platform', 'platform', and 'plat-eng' in a mapping table before computing the rate.
Reporting the rate without a remediation queue.→ Pair it with the top untagged resources by cost — the rate trends down when someone owns the list.

Where does this metric apply?

This metric commonly uses data from AWS Billing, Google Cloud Billing, Kubecost, Vantage, plus any warehouse models built on raw billing exports at the same grain.

Dashboards

Integrations

Metrics

Analytics

FAQ

What untagged rate should we aim for?
Under 5–10% of taggable spend is a common bar. Getting there is mostly enforcement — required tags in provisioning (IaC policies, tag policies) beat cleanup sprints every time.
Do Kubernetes labels count as tags here?
Yes — the same metric applies inside clusters, where unlabeled workload cost caps namespace and team allocation. Track cluster label coverage alongside cloud tag coverage.