Advisor Data Champion

An agent that champions data-driven decision making within organizations, helping teams establish data governance practices, quality standards, literacy programs, and analytics culture that turns raw data into reliable business intelligence.

When to Use This Agent

Choose Data Champion when:

• Establishing data governance frameworks and data ownership models
• Defining data quality standards and monitoring processes
• Building data literacy programs for non-technical stakeholders
• Creating data dictionaries and metadata documentation
• Advising on data strategy alignment with business objectives

Consider alternatives when:

• Building data pipelines (use a data engineering agent)
• Performing statistical analysis (use a data science agent)
• Designing database schemas (use a database architect agent)

Quick Start

Copy
# .claude/agents/advisor-data-champion.yml
name: Data Champion
model: claude-sonnet-4-20250514
tools:
  - Read
  - Write
  - Bash
  - Glob
  - Grep
prompt: |
  You are a data champion who advocates for data-driven decision
  making. Help organizations establish data governance, quality
  standards, and analytics culture. Focus on making data accessible,
  trustworthy, and actionable for all stakeholders.

Example invocation:

Copy
claude --agent advisor-data-champion "Help us establish a data
  governance framework for our growing startup. We have data
  in PostgreSQL, Snowflake, and various SaaS tools with no
  consistent definitions or ownership model."

Core Concepts

Data Governance Framework

Strategy → Standards → Ownership → Quality → Access → Compliance
   │          │           │           │         │          │
  Vision    Naming      Stewards   Monitoring  Policies  Regulations
  Goals     Definitions Councils   SLAs        Roles     Audit
  Metrics   Formats     Escalation Profiling   Catalog   Retention

Data Quality Dimensions

Dimension	Definition	Example Check
Accuracy	Data correctly represents reality	Email format validation
Completeness	Required fields are populated	Null rate < 5%
Consistency	Same value across systems	Customer name matches in CRM and billing
Timeliness	Data is current enough for use	Updated within last 24 hours
Uniqueness	No unintended duplicates	Deduplicated customer records
Validity	Data conforms to business rules	Age between 0 and 150

Data Maturity Model

Level	Characteristics	Focus Area
1 - Ad Hoc	No standards, tribal knowledge	Awareness and inventory
2 - Managed	Basic documentation, some owners	Governance framework
3 - Defined	Consistent standards, quality monitoring	Automation and tooling
4 - Measured	Quality SLAs, data contracts	Continuous improvement
5 - Optimized	Self-service analytics, data mesh	Innovation and democratization

Configuration

Parameter	Description	Default
maturity_target	Target maturity level	Level 3
governance_scope	Governance coverage	Critical datasets first
quality_monitoring	Quality check automation	Automated with alerts
catalog_tool	Data catalog platform	Open-source (DataHub)
ownership_model	Data ownership structure	Domain-based stewardship
compliance_reqs	Regulatory requirements	None specified
documentation_standard	Documentation template	Data dictionary format

Best Practices

1. Start with the most critical datasets, not all data. Trying to govern everything at once paralyzes the initiative. Identify the 5-10 datasets that drive the most important business decisions (revenue metrics, customer records, product usage data) and establish governance for those first. Success with critical data builds credibility and momentum for expanding to other datasets.

2. Assign data owners from the business side, not just IT. Data ownership belongs to the teams who generate and use the data. The marketing team should own marketing campaign data because they understand what the fields mean, what quality looks like, and how the data should be used. IT teams maintain the infrastructure; business teams define the standards and resolve quality issues.

3. Create a data dictionary with business definitions, not just technical schemas. A column called status could mean order status, account status, or approval status. The data dictionary should define each field in business language: "Order Status: The current fulfillment state of a customer order. Valid values: pending, processing, shipped, delivered, cancelled. Set by the fulfillment system when state changes." Technical stakeholders can read schemas; business stakeholders need human language.

4. Implement data quality monitoring as automated checks, not manual reviews. Define quality rules as code that runs automatically: null rate thresholds, valid value ranges, cross-system consistency checks, and freshness requirements. Alert data owners when quality drops below SLAs. Manual quality reviews catch issues days or weeks late; automated checks catch them within hours.

5. Measure data quality with business-meaningful metrics. Instead of tracking abstract quality scores, measure impact: "How many support tickets were caused by incorrect customer data? How many reports were delayed by missing data? What revenue was affected by stale pricing data?" Business-impact metrics get executive attention and budget that abstract quality scores never will.

Common Issues

Data governance initiative stalls due to lack of executive sponsorship. Governance requires cross-team coordination and sometimes conflicts with team autonomy. Without executive backing, teams can ignore governance standards. Secure a C-level sponsor by framing governance in business terms: regulatory risk, decision quality, and operational efficiency. A data incident (wrong report to the board, compliance violation) often creates the urgency that launches governance programs.

Teams resist data documentation because it feels like overhead. Make documentation part of the workflow, not a separate task. Integrate data dictionary updates into the pull request process for schema changes. Use tools that auto-generate technical documentation and only require business descriptions to be added manually. When documentation prevents a data incident or saves onboarding time, share that story to demonstrate value.

Data quality issues persist despite monitoring. Monitoring identifies problems; it doesn't fix them. Establish clear escalation paths: who gets notified? Who is responsible for fixing? What's the SLA for resolution? Without accountability, quality alerts become noise. Assign data quality incidents the same urgency and tracking as production incidents, with root cause analysis and prevention measures.