Data flow diagram

Updated
View as text
Open in Confluence
Open in ChatGPT Open in Claude

A data flow diagram answers a different question from a flowchart: not what happens in what order, but what data moves where. It has four elements only, which is why it stays readable when a flowchart of the same system would not.

Start at context level, one process and its external entities, then expand to level 1 only if you need to.

#When to reach for one

  • Privacy and data protection work, where the question is where personal data goes.

  • Integration design, before anybody has written a connector.

  • Explaining a system to someone who does not care about the code.

#How to draw it in Capable

Route

Use when

Graphviz

Layout is automatic, which suits diagrams that grow.

draw.io

You want the classic DFD shapes and full control of layout.

#A worked example

A level 1 data flow diagram for an order system:

Data flow diagram

The source, which you can paste into a new diagram and edit:

digraph dfd {
  rankdir=LR;
  node [fontname="Helvetica"];
  customer [shape=box, label="Customer"];
  warehouse [shape=box, label="Warehouse"];
  order [shape=ellipse, label="1.0\nTake order"];
  fulfil [shape=ellipse, label="2.0\nFulfil order"];
  orders [shape=box, style=dashed, label="D1 Orders"];
  customer -> order [label="order details"];
  order -> orders [label="new order"];
  orders -> fulfil [label="pending orders"];
  fulfil -> warehouse [label="pick list"];
  fulfil -> customer [label="confirmation"];
}

#A few things that catch people out

  • Every process needs at least one input and one output. A process with only inputs is a data store you mislabelled.

  • Do not draw control flow. Arrows carry data, not sequence.

  • Number processes (1.0, 1.1) so level 1 and level 2 diagrams can refer to each other.

Draw the unhappy path too.