Create Dashboard

2. Discover Existing Relevant Charts

Use mcp__Amplitude__get_charts to search broadly for existing charts before creating anything new. Search by:

• Product area name (e.g., "checkout", "onboarding", "notifications")
• Metric name (e.g., "DAU", "conversion", "retention")
• Event name (e.g., "purchase_completed", "signup")

For each relevant chart found, query it with mcp__Amplitude__query_charts to confirm it contains current, meaningful data. Prioritize reusing existing charts — they have proven definitions and are already trusted by the team.

Key principle: never create a chart if an equivalent one already exists. Only create new charts to fill genuine gaps.

3. Plan Dashboard Structure with Logical Sections

Organize charts into sections that tell a coherent story. Common section patterns:

• North Star: the single most important metric for the product area (1-2 charts)
• Acquisition funnel: how users enter and progress through key steps (funnel chart + segment breakdown)
• Engagement: depth and frequency of usage (retention, DAU/WAU/MAU, feature usage)
• Health and quality: errors, latency, support tickets, churn signals
• Experiments: active A/B tests and their results (if applicable)

Sequence sections from most strategic (top) to most operational (bottom). Executives read the top; engineers read the bottom.

4. Build and Organize

Call mcp__Amplitude__create_dashboard with:

• Dashboard name that clearly identifies the product area and audience
• Description explaining the dashboard's purpose
• Ordered list of chart IDs, grouped by section
• Layout configuration that gives more visual weight to North Star charts

For charts that need to be created (gaps identified in step 2), use the create-chart skill first, then add the resulting chart IDs to the dashboard.

5. Optimize Layout

After initial creation:

• Verify the top section loads with the most important metrics visible without scrolling
• Confirm chart titles are self-explanatory without requiring context (avoid abbreviations)
• Add text blocks between sections to explain what each section covers and why it matters
• Share the dashboard URL with the requester and confirm it meets their needs

MCP Tools

• mcp__Amplitude__get_charts — search for existing charts to reuse (call broadly, multiple searches)
• mcp__Amplitude__query_charts — validate that discovered charts contain current data
• mcp__Amplitude__create_dashboard — build the dashboard with organized chart layout
• mcp__Amplitude__get_context — get projectId and organization context (always call first)
• mcp__Amplitude__get_dashboard — load an existing dashboard to understand its structure before extending

Key Concepts

• North Star metric: The single most important metric that captures the core value the product delivers to users.
• Chart reuse: Prefer existing charts over new ones — they have established definitions and team trust.
• Dashboard section: A logical grouping of related charts within a dashboard, typically labeled by theme.
• Layout weight: Important charts should be displayed larger (full-width) than supporting charts.
• Decision-driven design: Every chart on a dashboard should exist because it informs a specific decision.
• Audience calibration: Executive dashboards emphasize outcomes (revenue, retention); engineering dashboards emphasize signals (errors, latency, event volume).

Output Format

The output includes:

1. A brief statement of the dashboard's purpose and intended audience
2. A list of sections with which charts are included in each, distinguishing between existing charts (reused) and new charts (created)
3. The dashboard URL in Amplitude
4. Any gaps that could not be filled because relevant events were not instrumented
5. Recommended next steps (e.g., add experiment results chart once the test launches, connect to a weekly email digest)