Data Trends>Data Visualization

What Is a Sankey Diagram? How to Read and Use One

key-takeawayKey Takeaways

  • A Sankey diagram shows how a measured quantity moves between categories or stages.
  • Link width represents relative flow amount when the chart uses proportional sizing.
  • Consider a Sankey when your question is how a measurable quantity moves, splits, combines, reaches a destination, or leaves a defined system.
  • Governed measures and consistent definitions make the visible flow more useful for analysis.
  • A wide link or drop-off points to a follow-up question; it does not prove why something happened.

When your analysis depends on tracing connected paths, a Sankey provides a flow-based view of the measured amounts between categories.

That visual pattern can help you see where to look next. To interpret it well, you still need a defined measure, consistent business terms, and a clear reporting scope.

What a Sankey Diagram Shows

A Sankey diagram is a quantitative flow visualization. Lawrence Livermore National Laboratory documents energy, water, and carbon flows as quantitative resource-flow examples. With proportional sizing, each band's width represents proportional flow quantity, as documented by Apache ECharts.

This design answers a specific kind of question: how does a measurable quantity move through a defined system? The quantity might be dollars, energy, records, shipments, people, or another numeric measure.

Before reading any band, check the chart's unit, period, and scope. A width has little meaning if you do not know whether it represents monthly revenue, annual shipments, or a percentage of a selected population.

Nodes, Links, and Measured Flow

Sankey diagrams have two main visual parts: nodes and links. Mermaid documentation describes categories or stages as nodes and their connections as links, and documents source, target, and value data as a common three-column pattern that can vary by platform.

Depending on the subject, a node might represent a source, process step, destination, or exit.

A link may split when one source feeds several destinations. Several links can also combine at a later node, which helps you see consolidation as well as distribution.

What about color? Unless the legend assigns a numeric meaning, treat color as a categorical aid that helps you follow related paths, not as a second measure.

How to Read a Sankey Diagram

A Sankey can look busy when many paths cross the page. Rather than scanning every band at once, use a repeatable sequence that starts with context and moves toward comparison.

First, identify what is flowing. Next, confirm the unit, reporting period, included population, and direction. Then trace a single path before comparing it with the rest of the chart.

The International Energy Agency explains how to read an energy Sankey by tracing supply through intermediate stages. Its example moves from supply on the left to final consumption or exit on the right. That arrangement is common, but it is not a rule for every Sankey.

Start with the Measure and Scope

Ask four questions before interpreting the shape:

  • What quantity is moving?

  • Which unit represents that quantity?

  • What time period does the chart cover?

  • Which categories, records, or populations are included?

Suppose two bands have similar widths. You cannot make a sound comparison if one reflects gross revenue and the other reflects net revenue, or if their reporting periods differ.

Scope matters for the same reason. If the chart includes only completed orders, it cannot tell you about every order that entered the process unless the excluded records are stated elsewhere.

Trace One Path Before Comparing the Whole Chart

Choose one source and follow its link through each connected stage. Note where the path splits, joins another path, narrows, or exits the displayed system.

Once you understand that path, compare it with an adjacent one. Are their widths based on the same measure, and do they cover the same period and population?

A wide band may identify a high-volume route. A narrow exit or visible drop may point to a place worth examining, but the U.S. Department of Education advises that additional research is needed to establish cause.

Use the diagram to frame the next question. You might ask which segments account for the visible difference, whether the pattern repeats over time, or whether a business rule changed during the period.

When a Sankey Diagram Fits Your Question

Consider a Sankey when your question centers on movement through connected stages. It can represent how a quantity splits, combines, reaches a destination, or leaves a defined system.

The chart is less useful when the connections do not matter. If you only need to rank five categories by value, a simpler display may communicate the answer with less visual work.

Questions a Sankey Can Help You Frame

A Sankey can support questions such as:

  • How does a budget move from funding sources to programs?

  • Where do records pass between process stages?

  • How do shipments move from origins to destinations?

  • Where do people continue, branch, or leave a defined journey?

  • How does an input divide among outputs and recorded losses?

These business examples are analogies for measurable flow analysis. Their value depends on the quality of the data, the category definitions, and whether the chart matches the decision at hand.

Phrase the question before you build the view. A focused prompt such as “How did completed orders move through fulfillment stages last quarter?” gives you clearer boundaries than “What is happening in operations?”

When Another Chart May Be Clearer

The Centers for Disease Control and Prevention describes bar charts for category comparisons and time-period comparisons as their primary use. Use one when you want to compare or rank categories without emphasizing connections. If time is the primary dimension, a line chart may be a better fit for showing change across regular intervals.

A network diagram can show relationships when there is no primary quantity moving between nodes. Meanwhile, a table may be the clearest choice when readers need exact values for a small set of categories.

There is no universal winner among these formats. Test the view with the people who will use it, especially when your Sankey contains many nodes or crossing links.

Build a Sankey Diagram with Reliable Flow Data

A polished diagram cannot repair an unclear measure. Before you draw the first link, define the quantity, prepare the flow records, and check that every stage uses the same business meaning.

One common build pattern uses a source, target, and value for each connection. Platforms vary, so confirm the required structure in the documentation for your chosen tool.

Structure Each Flow as Source, Target, and Value

This three-column pattern is a useful starting model, though other platforms may ask for additional fields or a different format.

SourceTargetValue
Website visitProduct view8,200
Product viewCart2,900
CartPurchase1,400

In this illustrative table, each row represents one connection. The value uses the same unit throughout, and the labels identify the stage at each end of the link.

Keep labels consistent across rows.

Govern the Measure Before You Draw the Flow

A useful flow begins with a shared definition. Document the metric formula, unit, reporting window, inclusion rules, category mapping, and owner before preparing the chart.

Why does this matter? If one team counts submitted orders while another counts paid orders, their flows may look comparable while answering different questions.

Governed measures give the chart a stable analytical base. They also support follow-up analysis because you can examine a segment without silently changing what the metric means.

This discipline belongs in a data leader’s analytics strategy, but it affects every practitioner who builds or reads the view. Clear definitions make collaboration easier when several teams rely on the same flow.

Validate Totals, Omissions, and Display Rules

Check the prepared data before charting it. Analyst Studio supports SQL, R, Python, data preparation, modeling, and chart building, which can help you inspect and validate the records behind an analysis.

Your validation pass should cover:

  • consistent units and reporting periods;

  • duplicate source-to-target records;

  • missing or unexpected categories;

  • stage totals that do not reconcile with the stated scope;

  • the selected tool’s documented handling of negative values and nulls;

  • category mappings that changed during the period.

Display rules need documentation too. If you group small flows, omit values below a threshold, normalize the data, or apply minimum ribbon widths, state that choice near the chart.

Those choices may improve readability, yet they also change what readers can infer from the view. A short note helps them distinguish a visual simplification from the underlying data.

How to Keep a Sankey Diagram Readable

A Sankey becomes hard to scan when it tries to answer several questions at once. The U.S. Web Design System advises you to focus on one main idea, so start with one analytical goal and include only the stages and paths needed to support it.

Readable does not mean stripped of context. Your job is to reduce visual noise while keeping the measure, scope, category logic, and meaningful exceptions visible.

Reduce Clutter Without Hiding Meaning

Group minor categories only when the combined category supports the question. If you use an “Other” node, state what it contains and which threshold placed items there.

You can also provide access to the detailed categories outside the main diagram. A companion table, filter, or drill path lets readers move from the summary to specific records without crowding the first view.

Be cautious with automatic limits. A tool that displays only the largest paths may remove a small but meaningful flow, particularly when the analysis concerns exceptions or risk.

When omissions are necessary, disclose them. Readers can then interpret the chart as a focused view rather than a complete map of every recorded path.

Use Color, Labels, and Ordering with Purpose

Use a consistent color system for related categories or stages. If color carries quantitative meaning, define that scale in the legend; otherwise, avoid suggesting that darker or brighter colors represent more flow.

Labels should name the category and, when space allows, show the value or percentage. Keep the unit visible so a reader does not need to hunt through surrounding text.

Arrange stages in a logical sequence, then validate the layout with the people who will use it.

Review the diagram at its intended publishing size and validate that the labels and paths remain usable for your audience.

Turn Flow Analysis into Your Next Question

A Sankey gives you a map of measured movement, but the map is only the beginning. The most useful next step is to investigate the segments behind a wide path, split, or exit while keeping the same governed definitions.

With interactive data visualizations, you can move from a summary view toward the records and segments that shape it. That exploration can help you ask whether a pattern is concentrated in a product, region, customer group, or reporting period.

ThoughtSpot is the Agentic Analytics Platform, and its enterprise analytics platform supports governed analysis across business questions. For a flow diagram, the practical goal stays simple: use the visible pattern to form a sharper question, then test that question against trusted data.

The Leader in Agentic Analytics

2026 Leader

2026 Leader

2026 G2 Enterprise

2026 G2 Enterprise

2026 G2 SMB

2026 G2 SMB

2026 G2 Momentum

2026 G2 Momentum