Uneven granularity makes complexity incomparable across tasks

What it is

Comparing two tasks’ complexity by step count, depth, or number of leaves is meaningful only when they have been split into the same kind of unit. Inconsistent task granularity is the state in which some branches are split to “click confirm” and others stop at “reconcile the inventory discrepancy,” so the counts compare the analyst’s patience, not the load of the work. Two trees of mixed grain cannot be said to differ in complexity, and a step-count gap cannot rank development priority.

Why it happens

A “step” has no natural size. Analysts split familiar operations finely and leave unfamiliar ones, or ones that live in someone else’s system, as a large black box. Fine branches contribute many leaves and look long; coarse branches contribute one leaf and look short, while that leaf may contain judgment, waiting, and coordination. If complexity is operationalized as node counts, grain size enters the measurement. Cross-task comparison also needs like stopping rules: one tree stopped at keystrokes and one at goals have incommensurable step gaps.

Studying it

Specify first what kind of unit a leaf should satisfy—for example “an action one person can complete without handing off, or a judgment that can be checked once”—with positive and negative examples. For tasks to be compared, check whether leaves fall in that type; split or roll up outliers until they do. When reporting complexity, report the grain rule and the type mix of leaves, not a step count alone. Run a sensitivity check at a second grain (counting only at the goal layer) and see whether the task ranking flips.

Where it stops holding

A single task may keep mixed grain on purpose, provided mixed counts are not used in a comparison—for example splitting a safety-critical branch deeper. Subjective difficulty, time, and error rates are another class of complexity evidence; they do not replace grain alignment, but they can be used when alignment is impossible. Splitting every task to keystrokes in the name of fairness violates a purpose-relative stopping rule, trading a wrong kind of evenness for useless detail.

Applying it

• Write the leaf-unit definition before comparing tasks, and audit both trees against it.
• Do not tell the business “this flow is more complex” from step counts while grain is unaligned.
• If comparison is needed and re-splitting cannot wait, report observed duration, failures, or number of judgment points, and declare the misalignment.
• Put the grain rule in the task-library template so new tasks join under the same rule rather than under one tree, one habit.

Related

• Same group: Q4.13.1 Decomposition needs a stopping rule; deeper is not automatically better · Q4.13.2 Mark sequential, parallel, and conditional relations among subtasks · Q4.13.3 The same hierarchy can serve training materials and interface design
• Adjacent: Q4.07 Task analysis · Q3.07 Task completion time
• Search terms: task granularity · comparable complexity · leaf unit