Foundations · 03 · 10 min

Tokens, how a process actually runs

This is the concept that separates people who draw BPMN from people who model it correctly.

Imagine a coin moving through the diagram. The specification calls it a token:

  • A start event creates a token.
  • The token travels along sequence flows.
  • An activity holds the token while the work happens, then passes it on.
  • A gateway splits or merges tokens by its own rules.
  • An end event consumes a token. When no tokens remain, the instance is finished.

Every modelling error in the “common mistakes” module is really a token error. Run all four scenarios below and watch what the gateways do.

Lab · Gateway scenariospick a scenario, then press Play
AND → AND: Both branches run, the join waits for both, one token continues. Correct.

Run it a few times before reading on. The two outcomes below are much more convincing once you have watched them happen.

Why this matters more than it looks

Both broken models look reasonable on a whiteboard. Neither is obviously wrong until you move a token through by hand. That is why the token rules are worth learning before the symbol catalogue: they are the thing that tells you a diagram is incorrect.

"Learn the token rules before the symbol catalogue. Nearly every diagram I am asked to review is wrong in a way that only shows up when you move a token through it by hand, and once you can do that, you never need anyone to tell you whether a model is correct."

After this module you can

  • Trace a token through a process by hand
  • Predict what a gateway does to the number of live tokens
  • Recognise a deadlock and token multiplication before they ship