Numbers in pink: mermaid's chart types

Use it in data work for: cost splits, storage by domain, ticket categories - any "what makes up the total" moment in a budget doc or postmortem. Above five or six slices, switch to a table; pies stop communicating.

pie showData title Nova warehouse spend
  "Compute" : 55
  "Storage" : 30
  "Egress" : 15

• Gotcha: values must be positive. Zero or negative values throw a parse error - filter them out before charting.
• Gotcha: labels need quotes, always - unquoted labels are the number-one pie syntax error.
• Order is yours: mermaid does not sort - list slices largest-first for the classic look.

Use it in data work for: prioritizing quality fixes, tool evaluations, model backlog triage - any effort-vs-impact conversation where a picture ends the debate faster than a spreadsheet.

quadrantChart
  title Nova data quality fixes
  x-axis Low effort --> High effort
  y-axis Low impact --> High impact
  quadrant-1 Plan properly
  quadrant-2 Do first
  quadrant-3 Quietly drop
  quadrant-4 Nice to have
  Null user ids: [0.2, 0.85]
  Late CRM loads: [0.45, 0.7]
  Duplicate events: [0.3, 0.5]
  Schema drift alerts: [0.8, 0.75]
  Emoji in names: [0.15, 0.1]

• Gotcha: quadrant numbering is fixed - quadrant-1 is top RIGHT, then counterclockwise. Mislabeling these flips your story.
• Gotcha: coordinates are 0 to 1, not percentages or raw values - normalize before plotting.
• Crowding: past ~8 points labels collide; shortlist before you chart.

Use it in data work for: volume trends, cost curves, row counts per load - the workhorse chart of every capacity review. Bar and line stack in one chart: bars for actuals, line for the target.

xychart-beta
  title "Nova pipeline volume by month"
  x-axis [Jan, Feb, Mar, Apr, May, Jun]
  y-axis "Events (millions)" 0 --> 40
  bar [12, 15, 19, 24, 30, 36]
  line [14, 18, 22, 26, 31, 38]

• Gotcha: series are positional. Values map to x-axis categories by ORDER - the fourth number is April, no keys, no gaps allowed.
• Gotcha: quote multi-word strings - the title and any category with a space need double quotes.
• No legend labels for series out of the box - say what bar and line mean in the caption or surrounding text.

Use it in data work for: pipeline volume accounting - where records enter, split, and land. A sankey makes data LOSS visible: that thin quarantine ribbon below is the chart equivalent of a failing quality check, and it often comes straight out of a GROUP BY.

sankey-beta
App events,Quality gate,60
CRM export,Quality gate,25
Quality gate,Warehouse,78
Quality gate,Quarantine,7
Warehouse,Dashboards,48
Warehouse,ML features,30

• Gotcha: it is real CSV. No pipes, no arrows - and a comma INSIDE a node name needs the whole name wrapped in double quotes.
• Gotcha: conservation is on you. Mermaid will happily draw inflows that do not match outflows - reviewers will notice. Make the sums balance (60 + 25 = 78 + 7).
• Node names ARE identity: the exact same string links flows together - "Warehouse" and "warehouse" become two nodes.

Use it in data work for: skill matrices, tool capability comparisons, maturity assessments - anything scored on the same scale across 4-6 dimensions. Two curves make gaps visible instantly.

radar-beta
  axis sql["SQL"], py["Python"], viz["Visualization"]
  axis ml["ML"], gov["Governance"]
  curve now["Team today"]{4, 3, 4, 2, 2}
  curve goal["Hiring target"]{5, 4, 4, 4, 3}
  max 5
  min 0

• Gotcha: curve values are positional - they map to axes in declaration order. Use the named form curve c{sql: 4, ml: 2} when order gets fragile.
• Gotcha: set max explicitly. Without it the scale is derived from the data, and two radars side by side silently use different scales.
• Keep it to 3-4 curves - beyond that the polygons overlap into noise.