Branch Metrics Reality Audit

This workflow turns “messy but persuasive” branch signals into decision-ready judgment. It starts by answering common questions from your knowledge base (definitions, metric owners, known data caveats). If the question isn’t covered, it routes the user into a guided menu of practical checks.

Operators and leaders can quickly choose the decision-shaped problem they’re facing—trusting branch numbers, spotting dirty signal, comparing branches, or deciding when automation is safe. Each path returns a crisp checklist and “what usually breaks first,” then loops back so the user can run another audit in the same conversation.

• Knowledge-base-first answers for fast, consistent definitions and known caveats
• Guided menu of decision-shaped audits (trust, dirty signal, attribution, comparisons)
• Practical red flags designed to catch bad data before the meeting becomes a confidence theater
• Clear guidance on when automation is reliable and when human judgment must stay in the loop
• Optional handoff to a human analyst for high-stakes or ambiguous cases

1. Trigger: A user starts the workflow from an inbound message.
2. Knowledge base check: The workflow attempts to answer using your knowledge base (metric definitions, data sources, branch rules, and known issues).
3. Menu prompt: If routing is needed, the user sees a button menu to pick what they’re trying to decide.
4. User chooses a path: The workflow routes based on the selected button:
   1. Which branch numbers deserve trust? Returns a quick trust audit (ownership, freshness, denominator sanity, and “too smooth to be true” checks), then returns to the menu.
   2. How to spot dirty signal early Returns early-warning signs (breaks in logging, missing segments, sudden step-changes, and suspicious consistency), then returns to the menu.
   3. Automation vs human judgment Explains when automation is safe (stable inputs, bounded impact, good monitoring) vs when you need humans (policy changes, distribution shifts, incentives), then returns to the menu.
   4. Messy evidence → usable insight Shows how to preserve truth while reducing noise (keep raw notes, track uncertainty, avoid over-cleaning away edge cases), then returns to the menu.
   5. Comparing branches & attribution Calls out common comparison traps (mix shifts, seasonality, channel spillover, selection bias), then returns to the menu.
   6. Build a signal culture Gives operating rules that reduce slide-making and increase decision-making (owners, pre-mortems, “what would change my mind”), then returns to the menu.
   7. Talk to an analyst Routes the conversation to human support.

• A WhatsApp-connected Calypso messaging channel where interactive buttons are supported
• Optional but recommended: a populated Calypso knowledge base with your metric definitions, owners, known data caveats, and branch measurement rules
• No additional credentials are required beyond your Calypso channel configuration