콘텐츠로 이동

title: Dashboard composition applies_to: [analytics, fintech, b2b-saas, all-data-ui] version: 1.0.0 last_updated: 2026-05 stability: stable

Dashboard composition

A dashboard is a single screen showing multiple metrics, charts, and tables. The composition (what goes where, how dense, how it scrolls) determines whether users understand their data or get lost.

Three dashboard archetypes

Type Use Density
Operational Real-time monitoring (DevOps, customer support, ATC) Very high
Analytical Exploration, reports, deep dives Medium-high
Strategic / Executive High-level KPIs, trends Low — few widgets, big numbers

Don't conflate them. An operational dashboard with 3 KPI cards is wasteful. An executive dashboard with 30 metrics is unreadable.

The information hierarchy

A dashboard reads in zones:

┌────────────────────────────────────────────────────────────┐
│ Title + filters + last updated                             │  ← chrome
├────────────────────────────────────────────────────────────┤
│  KPI 1     KPI 2     KPI 3     KPI 4                       │  ← top: hero stats
├────────────────────────────────────────────────────────────┤
│  [Primary chart — biggest, most important trend]           │  ← upper-mid: trend
├────────────────────────────────────────────────────────────┤
│  [Secondary chart] [Tertiary chart]                        │  ← lower-mid: comparisons
├────────────────────────────────────────────────────────────┤
│  [Detail table — filterable, sortable, paginated]          │  ← bottom: drill-in
└────────────────────────────────────────────────────────────┘

The order matches the 5-second test: at a glance, what should the user know?

  1. Top row (KPIs): the answer to "are we OK?" — single numbers, big, glanceable.
  2. Upper-mid (trend chart): the answer to "what's been happening?" — line/area chart over time.
  3. Lower-mid (comparisons): "what's driving it?" — breakdown by category, source, region.
  4. Bottom (detail table): "let me dig in" — sortable rows for analysts.

Top row: KPI cards

A row of Statistic components. See examples/component-statistic.md.

Count Use
2 Hero stat + comparison (revenue + growth)
3 Three independent KPIs
4 Standard executive dashboard (4-up grid)
5–6 Operational (more is worse but sometimes needed)
7+ Too many — group or move to filters

Each KPI card: - One number (the metric) - Label - Optional sparkline (mini line chart inline) - Optional delta indicator (vs prior period)

┌────────────────────────┐
│ 이번 달 매출             │
│ ₩2,847,500   ↑ 12%     │
│ ▁▂▃▅▆▇                  │
└────────────────────────┘

Primary chart placement

The main chart sits directly below the KPI row. It's: - Wide (full container width — 12 columns, or all of mobile width) - Tall enough to read trends (typically 240–360px) - Time-series almost always (line or area chart) - Shows the same metric as the hero KPI, but over time

If the hero KPI is "monthly revenue", the primary chart is "monthly revenue trend".

Secondary charts: 2-up

Two charts side by side. Common pairs: - Trend + Breakdown (line + pie/donut) - Categorical comparison (two bar charts) - Geographic + categorical (map + bar)

On mobile: stack vertically. Don't try to fit 2-up on phones.

Bottom: detail table

For users who want to drill in. See examples/component-table.md. - Filterable - Sortable - Paginated - Often: click row to open detail (pop-up or new screen)

The table is the escape hatch from "executive summary" to "I want the raw data".

Density decisions

Comfortable

For executive / strategic dashboards. Lots of whitespace, few widgets per screen, big numbers.

[KPI]  [KPI]              ← only 2 KPIs, big
   [primary chart]
[secondary]  [secondary]

Standard

The 4-up KPI + primary + 2-up secondary + table pattern above. Most dashboards.

Dense (operational)

Many widgets, smaller. Use sparingly. Consider layout grids:

┌────┐ ┌────┐ ┌────┐ ┌────┐ ┌────┐
│ KPI │ │ KPI │ │ KPI │ │ KPI │ │ KPI │
└────┘ └────┘ └────┘ └────┘ └────┘
[primary  full-width chart  ]
[chart] [chart] [chart]
[chart] [chart] [chart]
[detail table]

Operational dashboards are often single-page no-scroll — designed for wall-mounted displays or constant-attention screens. Different design language.

Filters and global controls

Top of the dashboard: chrome row with: - Time range picker: "오늘 / 7일 / 30일 / 90일 / 연도 / 사용자 지정" - Filter chips: applied filters with close buttons - Last updated indicator: "방금 업데이트됨" or "10분 전" - Refresh button (optional, for non-realtime dashboards) - Export / share (optional)

Filters apply to the whole page. URL-sync them so refresh + share preserves state.

┌──────────────────────────────────────────────────────────┐
│ Sales Dashboard                                           │
│                                                           │
│ [최근 30일 ▾]  [지역: 서울 ✕]  [카테고리 ✕]  · 방금 업데이트│
└──────────────────────────────────────────────────────────┘

Mobile patterns

Mobile dashboards are constrained by vertical screen. Approaches:

  1. Stack everything vertically — KPI cards full-width, charts full-width.
  2. Tabs at the top — "Overview / Sales / Users / Settings" — let user pick what to see.
  3. Cards-in-a-list — each section is a Card; user scrolls through.

For Korean consumer fintech apps (Toss / KakaoBank style): - Big hero stat at top (account balance with delta) - Quick action grid (송금, 결제, 가계부) - Recent activity list - Discover / promo cards

This is more "home screen" than "dashboard" — different pattern, but the principles overlap.

Color in dashboards

A dashboard with 5 charts can be visually chaotic. Restrain the palette:

  • One primary brand color for the main metric / chart fill.
  • One accent color for comparisons.
  • Money colors (--color-money-positive / --color-money-negative) for delta indicators.
  • Neutral grays for axes, gridlines, labels.

For categorical breakdown (5+ series): use a sequential or diverging palette per chart-color-encoding.md. Don't pick 5 random brand colors.

Refresh and freshness

Cite realtime-data.md for the deeper pattern. Quick rules: - Show "last updated" timestamp. - Auto-refresh every 30s–5min depending on data velocity. - For genuinely real-time: WebSocket updates with subtle "ping" indicator on changed values. - For batch-updated dashboards (every hour): static snapshot + "다음 업데이트: 14:00".

Empty / loading / error states

Each widget independently: - Loading: skeleton matching the widget's shape. - Empty: "데이터가 없습니다" + suggestion if applicable. - Error: "불러오기 실패" + retry — doesn't break other widgets.

Cite empty-states.md and error-states.md.

Density & responsive — rules of thumb

Container width Dashboard layout
< 600px (mobile) Single column, all widgets stacked, KPIs become a 2-up or scrollable horizontal row
600–960px (tablet) KPIs 4-up, primary full-width, secondary 1-up
960–1440px (laptop) Standard layout: 4 KPIs, primary, 2-up secondary, table
> 1440px (large desktop) Up to 6 KPIs, primary chart, 2 or 3-up secondary, side panel for filters

Print

Dashboards print badly by default. If print matters: - Hide nav, filters, header chrome. - Force 1-up layout (one widget per page). - Print "captured at" timestamp at the bottom.

Or: redirect to a PDF export endpoint that renders a print-optimized version server-side.

Accessibility

  • Use semantic landmarks: <main>, <section> per major area.
  • Each widget has a heading (<h2> or appropriate).
  • Charts: provide text/table alternative (toggle "show data table" link).
  • Color is not the only signal — pair with patterns or labels in charts.
  • Live region (aria-live="polite") for KPI value updates if real-time.
  • Keyboard reachable: every interactive widget tab-reachable.
  • See a11y/contrast.md for chart text/axis contrast requirements.

Korean considerations

  • "대시보드" or "현황" / "요약" depending on tone.
  • Date format: "2026.05.07" compact for axes.
  • Numerical units: "₩1.2M" or "₩120만" — pick one and be consistent. See money-and-amount.md.
  • Colors: Korean stock convention (red=up) applies for financial dashboards.
  • Time zone: KST always (UTC+9, no DST).

Common dashboard anti-patterns

  • Too many widgets — 12+ on one screen. User can't focus.
  • All charts the same color — no information hierarchy.
  • Inconsistent time ranges across widgets — KPI shows "this month", chart shows "last 30 days" → confusing.
  • No "last updated" indicator — user can't tell if data is fresh.
  • Animated number tickers as decoration — distracting, not informative.
  • Charts without titles or legends — ambiguous.
  • Chart and adjacent number disagreeing — different filters applied.
  • Dashboard is just a wall of tables — tables are detail; charts are summary.
  • Mobile dashboard that requires horizontal scroll — hostile.

Code example structure

function SalesDashboard() {
  const filters = useFilters();
  const data = useDashboardData(filters);

  return (
    <Page>
      <DashboardHeader>
        <PageTitle>매출 현황</PageTitle>
        <DashboardFilters value={filters} onChange={setFilters} />
        <LastUpdated timestamp={data.timestamp} />
      </DashboardHeader>

      {/* KPI row */}
      <KpiRow>
        <Statistic label="이번 달 매출" value={data.revenue} delta={data.revenueDelta} />
        <Statistic label="신규 가입" value={data.signups} delta={data.signupsDelta} />
        <Statistic label="활성 사용자" value={data.activeUsers} delta={data.usersDelta} />
        <Statistic label="평균 주문 금액" value={data.aov} delta={data.aovDelta} />
      </KpiRow>

      {/* Primary trend chart */}
      <ChartCard title="매출 추이" subtitle="지난 30일">
        <LineChart data={data.revenueTrend} />
      </ChartCard>

      {/* Secondary 2-up */}
      <Grid cols={2}>
        <ChartCard title="카테고리별 매출">
          <BarChart data={data.byCategory} />
        </ChartCard>
        <ChartCard title="지역별 매출">
          <PieChart data={data.byRegion} />
        </ChartCard>
      </Grid>

      {/* Detail table */}
      <Table data={data.transactions} columns={transactionColumns} pagination />
    </Page>
  );
}

Cross-reference