name: Keboola FinOps Guide
description: Financial operations guidance for managing Keboola platform costs. Keboola
  uses a usage-based pricing model tied to compute consumption (Project Power Units
  / Time Credits) and storage volume rather than per-seat licensing. Cost optimization
  focuses on right-sizing compute backends, scheduling jobs efficiently, and monitoring
  PPU consumption.
url: https://www.keboola.com/business-solutions/pay-as-you-go-pricing
created: '2026-06-13'
modified: '2026-06-13'
billing_model:
  type: Usage-based (compute + storage)
  description: Keboola charges based on compute time consumed (measured in Project
    Power Units or minutes) and data storage volume. There are no per-seat charges.
    Free tier users pay $0.14/minute for overage compute.
  units:
    - name: Project Power Units (PPUs) / Time Credits
      description: One PPU equals one hour of compute at the base (small SQL) rate.
        Higher-tier backends consume more PPUs per hour.
      enterprise: true
    - name: Compute Minutes
      description: Free tier unit. 60 minutes/month included; overages at $0.14/minute.
      free_tier: true
    - name: Storage GB
      description: Sum of table sizes in Keboola Storage. Free tier includes 250 GB.
cost_drivers:
  - driver: Compute Backend Size
    impact: high
    description: Larger compute backends (Medium, Large) consume 2-4x more PPUs per
      hour than Small. Right-sizing transformations to the minimum required backend
      is the primary cost lever.
    recommendation: Start with Small backends; scale up only when job runtime is unacceptably
      long. Profile transformation complexity before sizing.
  - driver: Job Frequency and Scheduling
    impact: high
    description: Every scheduled job consumes compute credits. Unnecessary frequent
      scheduling multiplies costs without data freshness benefit.
    recommendation: Set orchestration schedules aligned to actual data freshness requirements.
      Avoid sub-hourly schedules for non-critical pipelines.
  - driver: Data Source Connector Jobs
    impact: medium
    description: Connector (extractor) jobs are billed at 2 PPU/hour regardless of
      data volume transferred. Consolidating multiple small extractions reduces overhead.
    recommendation: Batch related data source jobs into single orchestrations where
      possible.
  - driver: Storage Volume
    impact: medium
    description: Tables accumulate in Storage; old snapshots and staging tables consume
      storage quota.
    recommendation: Implement retention policies. Delete temporary staging buckets
      after pipeline completion. Use table aliases instead of data copies where possible.
  - driver: DataApps and Sandboxes
    impact: medium
    description: Running DataApps and Python/R sandboxes consume PPUs while active.
      Long-running or idle workspaces can be significant cost drivers.
    recommendation: Configure automatic idle timeout for sandboxes. Shut down DataApps
      when not in active use.
  - driver: SQL Query Jobs (DWH Direct Query)
    impact: medium
    description: Ad-hoc queries executed via the Query API or DWH Direct Query consume
      8-32 PPU/hour depending on backend.
    recommendation: Use query result caching where available. Avoid exploratory queries
      against large tables without LIMIT clauses.
cost_optimization:
  strategies:
    - name: Right-size compute backends
      description: Match transformation backend size to actual computational needs.
        Use XSmall or Small for simple SQL transforms; reserve Large for complex aggregations
        over billions of rows.
    - name: Monitor PPU consumption in Management API
      description: Use the Management API to query project PPU usage programmatically.
        Set up alerts via the Notifications API when consumption exceeds thresholds.
    - name: Consolidate orchestrations
      description: Group related pipelines into single orchestrations to reduce per-job
        overhead and minimize idle time between sequential steps.
    - name: Archive or delete unused data
      description: Regularly audit Storage buckets and tables. Remove staging artifacts,
        expired exports, and legacy configurations to reduce storage billing.
    - name: Use linked buckets and aliases
      description: Share data across projects using linked buckets (which do not count
        toward storage limits) instead of duplicating tables.
    - name: Schedule during off-peak hours
      description: For Enterprise customers with dedicated compute, scheduling heavy
        transformations during off-peak hours can reduce contention and may lower
        costs on consumption-based contracts.
visibility_tools:
  - name: Keboola Management API
    description: Query project-level PPU consumption, storage usage, and job history
      programmatically.
    url: https://api.keboola.com/?service=manage
  - name: Keboola Billing API
    description: Access billing data and payment history for Pay-as-You-Go projects.
    url: https://api.keboola.com/?service=billing
  - name: Project Limits UI
    description: View current business and platform limits for a project in the Keboola
      web interface under project management settings.
    url: https://help.keboola.com/management/project/limits/
free_tier_management:
  description: Free tier users should carefully manage their 60 compute minutes/month
    to avoid unexpected $0.14/minute overage charges.
  tips:
    - Monitor remaining compute minutes in the project dashboard before running jobs
    - Disable scheduled orchestrations when not actively developing
    - Use the smallest compute backend (XSmall for Python/R) for development work
    - Delete test tables and buckets promptly to stay within 250 GB storage limit
    - Note that jobs are paused (not failed) when credits are exhausted; data is
      preserved