dashboard-from-research

name	dashboard-from-research
description	Transform research data (Excel, CSV, stickies) into structured insights and interactive dashboards

Dashboard from Research — Research Data to Interactive Dashboard

Transform research data (Excel, CSV, FigJam stickies, markdown) into structured insights and interactive dashboards with dataviz components. Chains research pipeline → specs → code generation → preview.

Freedom Level: High

Full autonomy over data interpretation, visualization choices, and dashboard layout. Must back every design decision with the research data.

When to Use

• User has research data (Excel, CSV, survey results, interview notes)
• FigJam board has stickies from workshops or brainstorming
• Need to create a dashboard that visualizes research findings
• Turning qualitative/quantitative data into actionable UI

Workflow

Step 1: Ingest Research Data

memi research from-file <path>     → Excel/CSV parsing
memi research from-stickies        → FigJam sticky notes
memi research synthesize           → Combine all sources

Output: research/insights.json with structured findings.

Step 2: Analyze & Categorize

Classify insights into dashboard-friendly categories:

Quantitative → KPI cards, charts, trend lines
  - Metrics: numeric values with labels
  - Time series: data over time → line/area charts
  - Comparisons: A vs B → bar charts
  - Distributions: spread → histograms

Qualitative → Text summaries, tag clouds, quotes
  - Themes: grouped findings → category cards
  - Quotes: user verbatims → quote components
  - Sentiment: positive/negative → sentiment indicators

Relational → Flow diagrams, matrices, maps
  - User journeys: step sequences → flow components
  - Relationships: connections → network graphs
  - Hierarchies: nested structures → tree views

Step 3: Create Specs (Atomic Design)

For each visualization need, create the right spec type:

KPI metric → memi spec component MetricCard (molecule)
  props: { title, value, change, trend, icon }

Trend chart → memi spec dataviz TrendChart
  chartType: "area" | "line"
  dataShape: { x: "date", y: "value", series: [...] }

Comparison → memi spec dataviz ComparisonChart
  chartType: "bar"
  dataShape: { category: "string", values: [...] }

The dashboard page → memi spec page ResearchDashboard
  layout: "dashboard"
  sections: [metrics-row, charts-row, insights-section, quotes]

Step 4: Generate Code

memi generate                      → all specs → React + Tailwind
memi preview                       → localhost preview server

Step 5: Design in Figma (Optional)

If the dashboard should also exist in Figma:

1. use_figma → create the dashboard layout using components
2. figma_take_screenshot → validate
3. Self-healing loop (max 3 rounds)
4. add_code_connect_map → establish design ↔ code parity

Dashboard Layout Pattern

Frame (VERTICAL, fill, 1280×900)
├── Header (HORIZONTAL, hug height, fill width, padding=24)
│   ├── Title: "Research Dashboard"
│   ├── Subtitle: research date range
│   └── Actions: export, filter, refresh
├── Metrics Row (HORIZONTAL, fill, gap=16, padding=24)
│   └── MetricCard × 4-6 (fill, equal width)
├── Charts Section (HORIZONTAL, fill, gap=16, padding=0-24)
│   ├── Primary Chart (2/3 width)
│   └── Secondary Chart (1/3 width)
├── Insights Grid (grid 2-3 col, gap=16, padding=24)
│   └── InsightCard × N
└── Detail Section (VERTICAL, fill, padding=24)
    └── DataTable or QuotesList

Data → Chart Type Decision

Data Pattern	Chart Type	Recharts Component
Single value + trend	KPI Card	Custom (Card + Badge)
Values over time	Area/Line	<AreaChart> / <LineChart>
Category comparison	Bar	<BarChart>
Part of whole	Pie/Donut	<PieChart>
Two dimensions	Scatter	<ScatterChart>
Distribution	Histogram	<BarChart> (binned)
Multiple metrics	Composed	<ComposedChart>

Anti-Patterns

• Creating charts without understanding the data first
• Using complex visualizations when a simple KPI card suffices
• Not including data source attribution
• Hardcoding sample data instead of connecting to research output
• Skipping the research synthesis step (going straight to UI)
• Not generating specs before code (violates spec-first)