Skip to content

Roadmap (high level)

This doc is intentionally short. It exists to set expectations.

Now (stability and trust)

  • streaming-first for big logs
  • structured logging and run correlation
  • reproducibility: config and dataset fingerprints

Next (power and sensitivity)

  • stronger experiment analysis features (variance reduction)
  • better segment regression surfacing
  • richer guardrails

Later (advanced evaluation)

  • more OPE diagnostics and ranking-specific estimators
  • interleaving variants and deeper analysis

Non-goals

  • training models
  • serving recommendations
  • replacing a full experimentation platform