Metrics Dashboard Builder

metrics-dashboard-builder

Agent: Data Analyst

L2 data analyst (Nx) responsible for data modelling, instrumentation implementation, metrics dashboards, funnel analysis, and signal synthesis.

Department ethos: ideal-data-growth.md

Skill Description

The metrics dashboard builder designs and implements dashboards that visualize product and business KPIs, enabling stakeholders to monitor health, detect regressions, and make data-informed decisions without writing SQL for recurring questions.

When to Use

• When a new product initiative defines success metrics that need ongoing monitoring.
• When stakeholders request recurring metrics that are currently answered by ad-hoc SQL queries.
• When a quarterly goal framework is finalized and each metric needs a tracking visualization.
• When an existing dashboard is outdated, slow, or no longer reflects current KPIs.

Workflow

1. Gather requirements: Identify the audience (executive, PM, engineering), the metrics to display, the required dimensions (time, cohort, segment), and the refresh cadence.
2. Design layout: Sketch the dashboard layout prioritizing the most critical metrics at the top. Group related metrics. Limit to 8-12 visualizations per dashboard to prevent cognitive overload.
3. Write queries: Build the SQL queries or BI tool calculations for each metric. Reference materialized views from the data model where available for performance.
4. Select chart types: Choose the appropriate visualization for each metric — time series for trends, bar charts for comparisons, scorecards for current values, funnels for conversion flows.
5. Add context: Include comparison baselines (prior period, target line), date filters, segment selectors, and metric definitions as tooltips or annotations.
6. Validate accuracy: Cross-check dashboard values against raw SQL queries for the same period. Confirm numbers match within acceptable rounding tolerance.
7. Publish and document: Deploy the dashboard, set access permissions, and write a one-page guide explaining each metric's definition, data source, and refresh cadence.

Anti-Patterns

• Dashboard sprawl: Creating a new dashboard for every request instead of extending existing ones fragments metrics across dozens of views. Why: stakeholders stop checking dashboards when they can't find the one they need.
• No metric definitions: Displaying numbers without explaining the calculation leads to conflicting interpretations. Why: "active users" means different things to different teams; the dashboard must disambiguate.
• Real-time when batch suffices: Building real-time dashboards for metrics reviewed weekly wastes engineering resources on unnecessary pipeline complexity. Why: the refresh cadence should match the decision cadence, not the technical capability.

Output

Success:

• A published dashboard with 8-12 visualizations covering the requested KPIs, comparison baselines, segment filters, and a documentation page with metric definitions and data sources.

Failure:

• Dashboard values do not match raw query results. Report the discrepancy, the affected metrics, the root cause (stale cache, wrong join, missing filter), and the correction applied.

Related Skills

• data-model-designer-data -- dashboards query the data model; schema changes may break dashboard queries.
• goal-framer-data -- the goal framework defines the metrics the dashboard must surface.
• alerting-configurator-data -- every dashboard metric should have corresponding alert coverage.
• metrics-dashboard-growth -- the growth dashboard focuses on acquisition/activation; this skill covers product and business KPIs broadly.