specificationVersion: '1.0'
id: evidently-finops
name: Evidently AI FinOps Framework
description: >
  FinOps guidance for managing and optimizing costs associated with Evidently AI deployments.
  Evidently AI follows a hybrid revenue model: the OSS library is free (Apache 2.0), while
  the commercial platform tiers (Expert, Enterprise) combine a base subscription with
  per-data-volume charges. Because the platform is self-hosted, total cost of ownership
  includes both Evidently licensing/subscription and operator infrastructure costs.
url: https://www.evidentlyai.com/pricing
focusVersion: '1.0'

costDimensions:
  - id: oss-infrastructure
    name: OSS Self-Hosted Infrastructure
    description: >
      Compute, storage, and networking costs for running the Evidently OSS platform.
      These are entirely borne by the operator and depend on the scale of evaluation
      jobs, number of traces stored, and dashboard query frequency.
    costType: Infrastructure
    billedBy: Operator (cloud/on-prem provider)
    examples:
      - Virtual machine or container compute for the Evidently server
      - Object storage for JSON report artifacts and raw trace data
      - Database storage for metrics time-series and test results
      - Network egress for dashboard access and API calls

  - id: subscription-fee
    name: Platform Subscription Fee
    description: >
      Base subscription for the Expert or Enterprise self-hosted platform tier. Starting
      at approximately $500/month for Expert. Enterprise is custom-quoted. Includes
      no-code workflows, RBAC, alerts, scheduled tasks, and managed support.
    costType: Subscription
    billedBy: Evidently AI
    billingCycle: Monthly or Annual
    startingPrice: 500
    currency: USD

  - id: data-volume-charges
    name: Per-Data-Volume Overage Charges
    description: >
      Commercial tiers include usage-based charges beyond the base subscription for
      data volume processed. Includes evaluation row counts, trace ingestion volume,
      and stored dataset size. Specific per-unit rates are not publicly published;
      obtain from Evidently sales.
    costType: Usage
    billedBy: Evidently AI
    unit: Data volume (rows / GB)
    rateNote: Not publicly disclosed; contact sales for details

optimizationStrategies:
  - id: oss-first
    name: Start with OSS Before Committing to Commercial Tier
    description: >
      The open-source library and self-hosted OSS platform provide the full evaluation
      and monitoring feature set at zero licensing cost. Teams should validate their
      use case and scale requirements with OSS before adopting the Expert tier.
    category: Cost Avoidance

  - id: batch-evaluations
    name: Batch Evaluations to Reduce Compute Costs
    description: >
      Evidently is designed for batch evaluation runs (e.g., nightly or per-release
      checks) rather than per-request real-time inference. Scheduling evaluations in
      off-peak windows reduces compute costs on the underlying infrastructure.
    category: Optimization

  - id: selective-metrics
    name: Use Only Required Metrics
    description: >
      With 100+ available metrics, computing all metrics for every run significantly
      increases CPU and time cost. Profile which metrics are actionable and limit
      evaluation scope to those, reducing per-run compute expenditure.
    category: Efficiency

  - id: startup-discount
    name: Apply for Startup Discount Program
    description: >
      Evidently AI offers eligible startups up to 90% discount on the Expert tier for
      six months via their startup program at https://www.evidentlyai.com/sign-up-startups.
      Qualifying teams should apply before committing to full pricing.
    category: Cost Reduction

  - id: self-host-vs-saas
    name: Self-Host to Avoid SaaS Markup
    description: >
      Since Evidently Cloud SaaS is discontinued, all commercial deployments are now
      self-hosted. Operators control infrastructure costs directly. Use spot/preemptible
      instances for batch evaluation workloads and right-size storage tiers based on
      retention policies for reports and traces.
    category: Architecture

budgetingGuidance:
  - category: Small Team (OSS)
    monthlyEstimate: "$0 licensing + $50–$200 infrastructure"
    notes: >
      Single developer or small team running batch evaluations on existing ML infrastructure.
      Storage and compute overhead is minimal.

  - category: Growing Team (Expert)
    monthlyEstimate: "$500+ subscription + $200–$1,000 infrastructure"
    notes: >
      Team of 5–20 with scheduled evaluations, LLM tracing, and dashboard access. Data
      volume overages may add to the base subscription cost.

  - category: Enterprise
    monthlyEstimate: "Custom quote + dedicated infrastructure"
    notes: >
      Large organizations with strict data governance. Budget includes Evidently Enterprise
      licensing, dedicated infrastructure (private cloud or on-prem), and implementation
      support costs.