Data Visualization
Create clear, effective data visualizations via inference.sh CLI.
Quick Start
Requires inference.sh CLI (belt). Install instructions
belt login
# Generate a chart with Python
belt app run infsh/python-executor --input '{
"code": "import matplotlib.pyplot as plt\nimport matplotlib\nmatplotlib.use(\"Agg\")\n\nmonths = [\"Jan\", \"Feb\", \"Mar\", \"Apr\", \"May\", \"Jun\"]\nrevenue = [42, 48, 55, 61, 72, 89]\n\nfig, ax = plt.subplots(figsize=(10, 6))\nax.bar(months, revenue, color=\"#3b82f6\", width=0.6)\nax.set_ylabel(\"Revenue ($K)\")\nax.set_title(\"Monthly Revenue Growth\", fontweight=\"bold\")\nfor i, v in enumerate(revenue):\n ax.text(i, v + 1, f\"${v}K\", ha=\"center\", fontweight=\"bold\")\nplt.tight_layout()\nplt.savefig(\"revenue.png\", dpi=150)\nprint(\"Saved\")"
}'
Chart Selection Guide
Which Chart for Which Data?
| Data Relationship | Best Chart | Never Use |
|---|
| Change over time | Line chart | Pie chart |
| Comparing categories | Bar chart (horizontal for many categories) | Line chart |
| Part of a whole | Stacked bar, treemap | Pie chart (controversial but: bar is always clearer) |
| Distribution | Histogram, box plot | Bar chart |
| Correlation | Scatter plot | Bar chart |
| Ranking | Horizontal bar chart | Vertical bar, pie |
| Geographic | Choropleth map | Bar chart |
| Composition over time | Stacked area chart | Multiple pie charts |
| Single metric | Big number (KPI card) | Any chart (overkill) |
| Flow / process | Sankey diagram | Bar chart |
The Pie Chart Problem
Pie charts are almost always the wrong choice:
❌ Pie chart problems:
- Hard to compare similar-sized slices
- Can't show more than 5-6 categories
- 3D pie charts are always wrong
- Impossible to read exact values
✅ Use instead:
- Horizontal bar chart (easy comparison)
- Stacked bar (part of whole)
- Treemap (hierarchical parts)
- Just a table (if precision matters)
Design Rules
Axes
| Rule | Why |
|---|
| Always start Y-axis at 0 (bar charts) | Prevents misleading visual |
| Line charts CAN start above 0 | When showing change, not absolute values |
| Label both axes | Reader shouldn't have to guess units |
| Remove unnecessary gridlines | Reduce visual noise |
| Use horizontal labels | Vertical text is hard to read |
| Sort bar charts by value | Don't use alphabetical order unless there's a reason |
Color
| Principle | Application |
|---|
| Max 5-7 colors per chart | More becomes unreadable |
| Highlight one thing | Grey everything else, color the focus |
| Sequential for magnitude | Light → dark for low → high |
| Diverging for positive/negative | Red ← neutral → blue |
| Categorical for groups | Distinct hues, similar brightness |
| Colorblind-safe | Avoid red/green only — add shapes or labels |
| Consistent meaning | If blue = revenue, keep it blue everywhere |
Good Color Palettes
# Sequential (low to high)
sequential = ["#eff6ff", "#bfdbfe", "#60a5fa", "#2563eb", "#1d4ed8"]
# Diverging (negative to positive)
diverging = ["#ef4444", "#f87171", "#d1d5db", "#34d399", "#10b981"]
# Categorical (distinct groups)
categorical = ["#3b82f6", "#f59e0b", "#10b981", "#8b5cf6", "#ef4444"]
# Colorblind-safe
cb_safe = ["#0077BB", "#33BBEE", "#009988", "#EE7733", "#CC3311"]
Text and Labels
| Element | Rule |
|---|
| Title | States the insight, not the data type. "Revenue doubled in Q2" not "Q2 Revenue Chart" |
| Annotations | Call out key data points directly on the chart |
| Legend | Avoid if possible — label directly on chart lines/bars |
| Font size | Minimum 12px, 14px+ for presentations |
| Number format | Use K, M, B for large numbers (42K not 42,000) |
| Data labels | Add to bars/points when exact values matter |
Chart Recipes
Line Chart (Time Series)
belt app run infsh/python-executor --input '{
"code": "import matplotlib.pyplot as plt\nimport matplotlib\nmatplotlib.use(\"Agg\")\n\nfig, ax = plt.subplots(figsize=(12, 6))\nfig.patch.set_facecolor(\"white\")\n\nmonths = [\"Jan\", \"Feb\", \"Mar\", \"Apr\", \"May\", \"Jun\", \"Jul\", \"Aug\", \"Sep\", \"Oct\", \"Nov\", \"Dec\"]\nthis_year = [120, 135, 148, 162, 178, 195, 210, 228, 245, 268, 290, 320]\nlast_year = [95, 102, 108, 115, 122, 130, 138, 145, 155, 165, 178, 190]\n\nax.plot(months, this_year, color=\"#3b82f6\", linewidth=2.5, marker=\"o\", markersize=6, label=\"2024\")\nax.plot(months, last_year, color=\"#94a3b8\", linewidth=2, linestyle=\"--\", label=\"2023\")\nax.fill_between(range(len(months)), last_year, this_year, alpha=0.1, color=\"#3b82f6\")\n\nax.annotate(\"$320K\", xy=(11, 320), fontsize=14, fontweight=\"bold\", color=\"#3b82f6\")\nax.annotate(\"$190K\", xy=(11, 190), fontsize=12, color=\"#94a3b8\")\n\nax.set_ylabel(\"Revenue ($K)\", fontsize=12)\nax.set_title(\"Revenue grew 68% year-over-year\", fontsize=16, fontweight=\"bold\")\nax.legend(fontsize=12)\nax.spines[\"top\"].set_visible(False)\nax.spines[\"right\"].set_visible(False)\nax.grid(axis=\"y\", alpha=0.3)\nplt.tight_layout()\nplt.savefig(\"line-chart.png\", dpi=150)\nprint(\"Saved\")"
}'
Horizontal Bar Chart (Comparison)
belt app run infsh/python-executor --input '{
"code": "import matplotlib.pyplot as plt\nimport matplotlib\nmatplotlib.use(\"Agg\")\n\nfig, ax = plt.subplots(figsize=(10, 6))\n\ncategories = [\"Email\", \"Social\", \"SEO\", \"Paid Ads\", \"Referral\", \"Direct\"]\nvalues = [12, 18, 35, 22, 8, 5]\ncolors = [\"#94a3b8\"] * len(values)\ncolors[2] = \"#3b82f6\" # Highlight the winner\n\n# Sort by value\nsorted_pairs = sorted(zip(values, categories, colors))\nvalues, categories, colors = zip(*sorted_pairs)\n\nax.barh(categories, values, color=colors, height=0.6)\nfor i, v in enumerate(values):\n ax.text(v + 0.5, i, f\"{v}%\", va=\"center\", fontsize=12, fontweight=\"bold\")\n\nax.set_xlabel(\"% of Total Traffic\", fontsize=12)\nax.set_title(\"SEO drives the most traffic\", fontsize=16, fontweight=\"bold\")\nax.spines[\"top\"].set_visible(False)\nax.spines[\"right\"].set_visible(False)\nplt.tight_layout()\nplt.savefig(\"bar-chart.png\", dpi=150)\nprint(\"Saved\")"
}'
KPI / Big Number Card
belt app run infsh/html-to-image --input '{
"html": "<div style=\"display:flex;gap:20px;padding:20px;background:white;font-family:system-ui\"><div style=\"background:#f8fafc;border:1px solid #e2e8f0;border-radius:12px;padding:24px;width:200px;text-align:center\"><p style=\"color:#64748b;font-size:14px;margin:0\">Monthly Revenue</p><p style=\"font-size:48px;font-weight:900;margin:8px 0;color:#1e293b\">$89K</p><p style=\"color:#22c55e;font-size:14px;margin:0\">↑ 23% vs last month</p></div><div style=\"background:#f8fafc;border:1px solid #e2e8f0;border-radius:12px;padding:24px;width:200px;text-align:center\"><p style=\"color:#64748b;font-size:14px;margin:0\">Active Users</p><p style=\"font-size:48px;font-weight:900;margin:8px 0;color:#1e293b\">12.4K</p><p style=\"color:#22c55e;font-size:14px;margin:0\">↑ 8% vs last month</p></div><div style=\"background:#f8fafc;border:1px solid #e2e8f0;border-radius:12px;padding:24px;width:200px;text-align:center\"><p style=\"color:#64748b;font-size:14px;margin:0\">Churn Rate</p><p style=\"font-size:48px;font-weight:900;margin:8px 0;color:#1e293b\">2.1%</p><p style=\"color:#ef4444;font-size:14px;margin:0\">↑ 0.3% vs last month</p></div></div>"
}'
Heatmap
belt app run infsh/python-executor --input '{
"code": "import matplotlib.pyplot as plt\nimport numpy as np\nimport matplotlib\nmatplotlib.use(\"Agg\")\n\nfig, ax = plt.subplots(figsize=(10, 6))\n\ndays = [\"Mon\", \"Tue\", \"Wed\", \"Thu\", \"Fri\", \"Sat\", \"Sun\"]\nhours = [\"9AM\", \"10AM\", \"11AM\", \"12PM\", \"1PM\", \"2PM\", \"3PM\", \"4PM\", \"5PM\"]\ndata = np.random.randint(10, 100, size=(len(hours), len(days)))\ndata[2][1] = 95 # Tuesday 11AM peak\ndata[2][3] = 88 # Thursday 11AM\n\nim = ax.imshow(data, cmap=\"Blues\", aspect=\"auto\")\nax.set_xticks(range(len(days)))\nax.set_yticks(range(len(hours)))\nax.set_xticklabels(days, fontsize=12)\nax.set_yticklabels(hours, fontsize=12)\n\nfor i in range(len(hours)):\n for j in range(len(days)):\n color = \"white\" if data[i][j] > 60 else \"black\"\n ax.text(j, i, data[i][j], ha=\"center\", va=\"center\", fontsize=10, color=color)\n\nax.set_title(\"Website Traffic by Day & Hour\", fontsize=16, fontweight=\"bold\")\nplt.colorbar(im, label=\"Visitors\")\nplt.tight_layout()\nplt.savefig(\"heatmap.png\", dpi=150)\nprint(\"Saved\")"
}'
Storytelling with Data
The Narrative Arc
| Step | What to Do | Example |
|---|
| 1. Context | Set up what the reader needs to know | "We track customer acquisition cost monthly" |
| 2. Tension | Show the problem or change | "CAC increased 40% in Q3" |
| 3. Resolution | Show the insight or solution | "But LTV increased 80%, so unit economics improved" |
Title as Insight
❌ Descriptive titles (what the chart shows):
"Q3 Revenue by Product Line"
"Monthly Active Users 2024"
"Customer Satisfaction Survey Results"
✅ Insight titles (what the chart means):
"Enterprise product drives 70% of revenue growth"
"User growth accelerated after the free tier launch"
"Support response time is the #1 satisfaction driver"
Annotation Techniques
| Technique | When to Use |
|---|
| Call-out label | Highlight a specific data point ("Peak: 320K") |
| Reference line | Show target/benchmark ("Goal: 100K") |
| Shaded region | Mark a time period ("Product launch window") |
| Arrow + text | Draw attention to trend change |
| Before/after line | Show impact of an event |
Dark Mode Charts
belt app run infsh/python-executor --input '{
"code": "import matplotlib.pyplot as plt\nimport matplotlib\nmatplotlib.use(\"Agg\")\n\n# Dark theme\nplt.rcParams.update({\n \"figure.facecolor\": \"#0f172a\",\n \"axes.facecolor\": \"#0f172a\",\n \"axes.edgecolor\": \"#334155\",\n \"axes.labelcolor\": \"white\",\n \"text.color\": \"white\",\n \"xtick.color\": \"white\",\n \"ytick.color\": \"white\",\n \"grid.color\": \"#1e293b\"\n})\n\nfig, ax = plt.subplots(figsize=(12, 6))\nmonths = [\"Jan\", \"Feb\", \"Mar\", \"Apr\", \"May\", \"Jun\"]\nvalues = [45, 52, 58, 72, 85, 98]\n\nax.plot(months, values, color=\"#818cf8\", linewidth=3, marker=\"o\", markersize=8)\nax.fill_between(range(len(months)), values, alpha=0.15, color=\"#818cf8\")\nax.set_title(\"MRR Growth: On track for $100K\", fontsize=18, fontweight=\"bold\")\nax.set_ylabel(\"MRR ($K)\", fontsize=13)\nax.spines[\"top\"].set_visible(False)\nax.spines[\"right\"].set_visible(False)\nax.grid(axis=\"y\", alpha=0.2)\n\nfor i, v in enumerate(values):\n ax.annotate(f\"${v}K\", (i, v), textcoords=\"offset points\", xytext=(0, 12), ha=\"center\", fontsize=11, fontweight=\"bold\")\n\nplt.tight_layout()\nplt.savefig(\"dark-chart.png\", dpi=150, facecolor=\"#0f172a\")\nprint(\"Saved\")"
}'
Common Mistakes
| Mistake | Problem | Fix |
|---|
| Pie charts | Hard to compare, always misleading | Use bar charts or treemaps |
| Y-axis not starting at 0 (bar charts) | Exaggerates differences | Start at 0 for bars, OK to truncate for lines |
| Too many colors | Visual noise, confusing | Max 5-7 colors, highlight only what matters |
| No title or generic title | Reader doesn't know the insight | Title = the takeaway, not the data type |
| 3D charts | Distorts data, looks unprofessional | Always use 2D |
| Dual Y-axes | Misleading, hard to read | Use two separate charts |
| Alphabetical sort on bar charts | Hides the story | Sort by value (largest first) |
| No labels on axes | Reader can't interpret | Always label with units |
| Chartjunk (decorative elements) | Distracts from data | Remove everything that doesn't convey information |
| Red/green only for color coding | Colorblind users can't read | Use shapes, patterns, or colorblind-safe palettes |
Related Skills
npx skills add inference-sh/skills@pitch-deck-visuals
npx skills add inference-sh/skills@technical-blog-writing
npx skills add inference-sh/skills@competitor-teardown
Browse all apps: belt app list
