Sankey Diagram Generator for Research Flows

What does this Sankey diagram generator do?

A Sankey diagram shows how a quantity moves between states or stages. Nodes are drawn as bars and flows as ribbons whose width is proportional to the amount. This Sankey diagram generator reads a simple source, target, and value table and does the bookkeeping before any drawing happens. It sums duplicate links, computes each node's inflow, outflow, and throughput, and assigns nodes to columns. It also calculates each flow's share of its source. The image model receives those numbers as fixed values, so ribbon widths, node totals, and percentages reflect your data instead of a visual guess. It works as a Sankey diagram maker for clinical pathways, energy and material flow analysis, carbon or nutrient budgets, and cohort studies.

How nodes, stages, and flows are computed

• Throughput is the larger of a node's inflow and outflow, and it sets the node height and the total in its label.
• Columns follow the longest path from the source nodes, so every ribbon moves forward. Add "align end nodes right" to push final outcomes into the last column.
• Within a column, nodes are ordered by throughput, largest first. Write "keep source order" to use the order in which nodes first appear in your file.
• Each link reports its percent of the source node's outflow, which is useful for pathway proportions such as "38% of ED patients were admitted".
• Diagrams are capped at 30 nodes and 60 links. Past that, the smallest nodes in a column are merged into an "Other" node, and a note lists what was merged.

How to read a Sankey chart

• Read from left to right: each column is a stage, and ribbon width shows how much flows between two nodes.
• Compare node heights to see where most of the quantity ends up, and ribbon widths to see the dominant routes.
• When an intermediate node receives more than it passes on, the difference is storage, loss, or missing follow-up. The generator flags these imbalances instead of hiding them.
• Colors usually track the source node, so you can follow one origin through later stages.

Tips and common mistakes

• Sankey diagrams need a directed acyclic flow. A loop such as Ward → ICU → Ward produces a clear error that names the cycle. Rename repeated stages, for example "Ward (post-ICU)".
• A node cannot flow into itself, and flows must be positive. Rows with zero flow are ignored.
• Keep one unit throughout, and state it in your instructions ("units: MWh") so it appears in labels.
• Values on links are printed only when you ask ("print values on links" or "show percent on links"). Node totals are always shown.
• For budgets, add explicit loss or storage nodes so inputs and outputs balance and readers can see the gap.

When to use a flow diagram, and when not to

Choose a Sankey chart when the story is about magnitude along routes: where patients go, how energy is converted and lost, or how a cohort splits over time. If you only need composition at a few time points, a stacked bar chart is easier to compare precisely. If flows are truly bidirectional or cyclic, such as migration between regions, a chord diagram or transition matrix heatmap is a better fit.