docs: address documentation gaps from three-year client support audit

[bertilhatt]

Motivation

I used Claude to go through Slack channel with clients and identify patterns of responses and product information mentioned there. I also used it to match with our docs, compare and identify topics that were not covered in the docs. It suggested this (very long) PR.

Summary

Adds callouts, warnings, and new sections across 25 documentation pages to address gaps identified during a systematic three-year audit of ~75 client Slack channels (430+ findings documented in the internal playbook audit). Every change addresses a pattern where clients repeatedly needed information that wasn't in the public docs.

Changes are organized by client impact:

High priority (7 items — frequent questions, high confusion):

• Run Log as debugging tool — new section in Debugging Metrics explaining when and how to use the Run Log
• Diagnostic query approximation caveat — callout on Diagnostics page noting queries exclude CUPED++, winsorization, and mixed-assignment filtering
• SSC does not account for CUPED++ — notes on both Sample Size Calculator and Progress Bar pages
• Entry point replaces assignment timestamp — warning callout on the entry point filter page
• Assignment logger required fields — callout on Assignment Logging page highlighting that omitting experiment causes silent data loss
• Backfill decision tree — new subsection in Data Pipeline explaining when a full refresh is (and isn't) needed
• Ramp via passthrough vs variant weight change — new section in Running Well-Powered Experiments

Medium priority (11 items — recurring confusion):

• ConfigurationStore environment isolation warning (iOS + Android SDK init)
• Holdout limitations: archived bands still consume hash space, compounding percentages, immutable dates, Admin-only creation
• Flag archival is irreversible
• JSON variation int-to-double type coercion
• Layer parameter lock on active experiments
• CDN stale-if-error does not cover 429 responses
• EPPO_OUTPUT schema accumulation guidance
• Service account rotation checklist (Snowflake, Redshift, BigQuery)
• Winsorization truncation (capping) vs discard clarification
• Schedule limitations (no per-day-of-week, reassignment via Metrics tab)

Lower priority (7 items):

• SSO is SP-initiated only (IdP-initiated causes login loop)
• Audience builder: no "is null" condition, no array-type attributes
• Interaction effects limited to Simple metrics
• Sequential test three-trigger readiness criteria
• Global Lift non-additivity of unique entity metrics
• Experiment key non-uniqueness note

3 lower-priority items were skipped as out of scope for this repo (billing model, API status field in Swagger, incident communication runbook).

Test plan

• Verify docs build successfully (Docusaurus)
• Spot-check rendered callout boxes on key pages (diagnostics, assignment-logging, holdout-config)
• Review that internal links resolve correctly (e.g., diagnostics anchor link from confidence-intervals page)

Made with Cursor

[netlify]

✅ Deploy Preview for eppo-data-docs ready!

Name	Link
🔨 Latest commit	ae771d9
🔍 Latest deploy log	https://app.netlify.com/projects/eppo-data-docs/deploys/69beed9de73cda00084d3633
😎 Deploy Preview	https://deploy-preview-721--eppo-data-docs.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.