BW-301h · Module 2

Decision Trees and Step Lists

4 min read

Every process contains two types of moments: moments where the executor makes a judgment and moments where the executor performs a specific action. Decision trees handle the judgment moments. Step lists handle the action moments. Conflating the two — trying to write a step list for a judgment call, or a decision tree for a mechanical sequence — produces a document that is neither navigable nor instructive. The architecture separates these two types of content and presents each in its appropriate format.

Do This

  • Use decision trees for judgment calls — "if X, go to step A; if Y, go to step B; if neither, escalate"
  • Use numbered step lists for sequential, deterministic actions — each step is discrete, actionable, and produces a specific output
  • Place the decision tree before the relevant step lists so the executor knows which path to take before beginning the steps
  • Keep step lists to a maximum of seven steps — anything longer is either two processes or a process that needs to be redesigned

Avoid This

  • Write judgment calls as prose that the executor must interpret — "use your best judgment based on the client's tone and history" in a step list is not a step
  • Embed a decision within a step list using conditional language — "in step 4, if the customer has not been contacted, first do X" creates a branch that belongs in a decision tree
  • Write a step list for a process with significant variability — variable processes need decision trees, not step lists that pretend the variability does not exist
  • Number more than seven steps without questioning whether the process should be broken into sub-processes