A finer break point comes around sooner, but it shields a task less than a coarser one does

What it is

Waiting for a good moment to interrupt runs straight into a tradeoff: a fine-grained boundary — one step inside a subtask finishing — comes around every few seconds, while a coarse-grained boundary — a whole subtask or phase finishing — might not show up for minutes. Waiting for the fast-arriving fine boundary means accepting a bigger disruption when you do interrupt; waiting for the slow-arriving coarse boundary means the pending information arrives later. This is not a question of which boundary is "better" — it is a tradeoff between latency and protection.

Why it happens

A fine boundary closes only the innermost sub-goal on the stack; the outer subtask's goal state is still hanging in working memory, unclosed. Interrupting there still forces the user to rebuild that outer goal state — only slightly less costly than interrupting at a completely arbitrary moment. A coarse boundary closes more layers of the stack, leaving less pending state to maintain, so the interruption costs less. But the deeper the goal hierarchy and the more steps it takes to clear it, the longer the natural gap between coarse boundaries. Protection and frequency are two opposite quantities produced by the same hierarchical structure: the coarser the grain, the slower it arrives; the finer the grain, the faster.

Studying it

Building on the same hierarchical decomposition, this kind of study logs how often boundaries of each grain actually occur in real usage traces, and pairs that with the measured resumption cost (post-interruption response-time increment, error rate) of interrupting at each grain — plotting a correspondence between "how often this boundary shows up" and "how much it costs to interrupt there."

Common independent variable: which goal-hierarchy level the boundary sits at. Common dependent variables: the average inter-arrival interval of that grain's boundaries, and the interruption-cost metric measured at that grain.

Methodological caveat: inter-arrival intervals depend heavily on the specific task and an individual's own pacing, so the same grain can vary widely across users and task instances. Report the distribution alongside the mean — a single average is a poor stand-in for "how long you'll typically wait."

Where it stops holding

• The tradeoff only matters when a task genuinely has boundaries at multiple grains and the pending information can actually afford to wait. If only one grain is distinguishable, or the information must be delivered instantly, the tradeoff doesn't apply.
• The shallower a task's hierarchy, the smaller the latency gap between coarse and fine boundaries, and the smaller the practical stakes of the tradeoff; the deeper the hierarchy, the more the tradeoff bites.
• This entry does not cover how to actually schedule notifications around this tradeoff — that is a separate, deferral-strategy question.

Applying it

• For deferrable, low-value-per-item information, lean toward waiting for the coarser boundary even if it means a longer wait; for time-sensitive content whose value itself decays with delay, shift to the finer boundary or deliver immediately, since a long wait can make the information worthless anyway.
• Do not apply one fixed "wait window" across every scenario. Boundary intervals differ by task depth — a fixed wait time either drags on too long for shallow tasks or never reaches a genuinely clean boundary in deep ones.
• Verification: log the real inter-arrival distribution of coarse and fine boundaries for the target task, then overlay the maximum tolerable wait for information at different urgency levels. Where the two curves cross tells you which grain to target — not a threshold set by intuition.

Related

• Same group: A5.14.1 Breakpoints come in coarse and fine grain, and the subtask boundary is only one coarser level · A5.14.3 Batching notifications to the next boundary costs less overall than immediate delivery · A5.14.4 User-set do-not-disturb windows compensate for limited automatic boundary detection
• Adjacent: A5.08 Interruption cost and task resumption
• Search terms: fine-grained breakpoint · boundary latency tradeoff · opportune moment for interruption