Use Sankey for flows between stages or systems, keep labels concise, avoid unreadably dense networks, and verify the final SVG before publishing.
Review the source structure, chart-specific interpretation risks and a copyable accessible summary. The chart itself remains the primary workspace above.
Reviewing the current data…
This review checks structure and common interpretation risks; it does not prove that the source data, measurement method, denominator, units or conclusion are valid.
Inspect flow structure numerically so missing branches, unintended cycles and large node imbalances are visible before export.
| Node | Inflow | Outflow | Net | Role |
|---|
Start with Graph Maker or CSV Chart Maker, then use a focused chart type when the visual question changes. Keep source data and export evidence connected instead of treating a rendered picture as the only record.
Sankey diagrams show entered source→target flows with proportional widths.
Review self-links, duplicate edges and cycles before publishing; the entered flow total is not automatically a conserved system total.
Use the product workflow above first. These notes add interpretation, verification and limits below the interactive result area.
A Sankey chart is easiest to trust when every source-target-value row has a clear unit and the major nodes approximately reconcile. Use the balance audit to find nodes where incoming and outgoing totals differ before spending time on colors or labels.
Band thickness only has meaning when values are comparable. Mixing dollars, percentages and counts in the same connected diagram can make a visually impressive chart that is quantitatively meaningless. Convert values to one unit or split the story into separate diagrams.
Dozens of very small links can turn labels and bands into visual noise. Combine genuinely similar minor categories into an Other group when the analytical question allows it, but keep the source data available so the aggregation is transparent and reversible.
A source-target table can contain a path that eventually returns to an earlier node. That may be legitimate for recirculation or repeated process stages, but it changes how totals should be interpreted. Use the cycle audit to distinguish a true process loop from an accidental data-entry reversal.
Before publishing, compare a few large band values and node totals with the pasted source rows and keep an accessible text summary with the graphic. The chart communicates structure; the table remains the easiest way to audit exact values and units.
WebToolArc