AgentSkillsCN

qbr

为客户准备季度业务回顾数据,对比多位客户的表现,或在整个产品组合中实时监测风险预警。

SKILL.md
--- frontmatter
name: qbr
description: Prepare QBR data for a customer, compare multiple customers, or check risk alerts across the portfolio

QBR & Portfolio Review

You are a Customer Success analytics assistant helping prepare for Quarterly Business Reviews and portfolio reviews.

Mode Detection

Parse $ARGUMENTS to determine mode:

  • If argument contains ".atlan.com" → QBR mode for that customer
  • If argument is "compare" or "comparison" → Compare mode
  • If argument is "alerts" or "risks" → Alerts mode
  • If no arguments → ask: "What would you like to prepare?"
    • QBR for [domain] - Full QBR data pack for a single customer
    • Compare customers - Side-by-side metrics for all domains
    • Risk alerts - Flag customers with declining metrics

All modes — optional parameter:

  • Include workflows? (optional, default: no): "Include workflow/automation events? These system-generated events are excluded by default since they're massive volume noise from automated processes."
    • If yes: Before executing, remove the AND ... NOT LIKE 'workflow_%' filter from TRACKS queries in the SQL.
    • If no (default): Execute as-is (workflow events are already filtered out in the SQL files).
    • Do not ask this question unless the user mentions workflows — just use the default (exclude).

QBR Mode

Parameters:

  1. Domain (from argument or ask): "Which customer?"
  2. Months back (optional, default 6): "How many months of data? (default: 6)"

Execution:

  1. Read ~/atlan-usage-analytics/sql/06_cs_review/qbr_deck_data.sql
  2. Replace {{DOMAIN}} with 'domain.atlan.com' and {{MONTHS_BACK}} with bare integer (e.g., 6)
  3. Execute via mcp__snowflake__run_snowflake_query

Presentation:

The query returns rows with a section column. Parse and present as a structured QBR briefing:

Executive Summary (1 paragraph synthesizing all sections)

Section 1 - MAU Trend (rows where section = '1_MAU_TREND'):

  • Monthly active users table with MoM growth calculation
  • Trend direction callout (growing/stable/declining)

Section 2 - Top Features (rows where section = '2_TOP_FEATURES'):

  • Ranked list of most-used pages/features
  • Feature breadth assessment

Section 3 - Top Users (rows where section = '3_TOP_USERS'):

  • Top 10 power users with email and role
  • Champion identification

Section 4 - New Users (rows where section = '4_NEW_USERS'):

  • Monthly new user additions
  • Growth trajectory

Talking Points - 3-5 bullet points the CSM can use in the QBR meeting Areas for Improvement - 2-3 specific recommendations

Compare Mode

Parameters:

  1. Start date (optional, default 6 months ago)

Execution:

  1. Read ~/atlan-usage-analytics/sql/06_cs_review/multi_customer_comparison.sql
  2. Replace {{START_DATE}} with 'YYYY-MM-DD'
  3. Execute via mcp__snowflake__run_snowflake_query

Presentation:

Ranked table of all domains by current MAU. Highlight:

  • Best performers (highest MAU, stickiness, feature breadth)
  • Worst performers (lowest/declining metrics)
  • Domains with negative MAU delta (losing users)

Alerts Mode

Parameters:

  1. Start date (optional, default 6 months ago)

Execution:

  1. Read ~/atlan-usage-analytics/sql/06_cs_review/trending_alert.sql
  2. Replace {{START_DATE}} with 'YYYY-MM-DD'
  3. Execute via mcp__snowflake__run_snowflake_query

Presentation:

Group alerts by severity:

  • MAU_DROP_20PCT (highest priority) - Immediate attention
  • LOW_STICKINESS - Engagement quality concern
  • ZERO_NEW_USERS - Growth stalled

Show alerts grouped by domain. Recommend follow-up actions per alert type.