Low-value notices should not interrupt

What it is

A notice has two roads: interrupt (sound, banner, screen takeover) and inform (silent shade, a mark, an inbox). Low value means missing it has almost no irreversible loss—achievements, like-digests, feature pitches, state sync for something already seen in-app. Those items may exist; they should not take the interrupt channel. Even at a cheap moment, cutting a low-value item into an ongoing task still bills resumption cost for near-zero gain.

This is channel choice: is an interrupt worth it. It is not how urgency is graded, and not whether volume will get the whole app disabled.

Why it happens

The interrupt channel is scarce. Every cut-in spends credit that the next item will still be taken seriously. Low-value items spend that credit and move the prior toward "probably nothing"; later, when an interrupt is warranted, people have learned to delay or swipe away. The inform channel does not steal working memory; people can sweep it at their own boundary. Leaving low value there is not concealment. It is refusing to buy optional information with resumption lag.

"Low value" is the recipient's loss, not the sender's expected clicks. A promotion with a high open rate can still be a low-value interrupt: opens follow color and countdown, not miss-cost.

Studying it

Assign the same low-consequence items at random to banner-plus-sound versus shade-only, during a real primary task. Compare primary-task errors, resumption time, and later memory for the item.

Independent variables: channel (interrupt / inform), miss-cost of the item, load of the primary task. Dependent variables: resumption lag, recall of the information, later delay in opening similar items, dismiss gestures in-session.

Lab instructions to "please look at every notice" promote low-value items into the task itself. Allow ignoring. Open rate is not success: interrupts can raise opens and primary-task errors together—that is buying clicks with mistakes.

Where it stops holding

For people who use the app as a workbench, ops messages can be redefined as work (community managers need every interaction). "Low value" must be rewritable by role, but that rewrite cannot become the default for everyone. If an assistive-tech user can learn about new information only through the interrupt channel (the shade is hard to scan), full silence withholds informing; they need a path that does not steal focus yet is still announced. On idle lock screens interrupt cost is low, but training "lock screen may sound" for junk spills into high-load contexts.

Applying it

• Label every type with a channel: interrupt, silent shade, in-app only. Types with low miss-cost must not use sound or occupying banners.
• Do not interrupt for state the person already saw in the foreground; at most leave a tappable record in the shade.
• Treat non-opens as an informing failure to fix with copy and placement, not as a reason to add a ring.
• Verify by taking a type with decent opens but no later action (achievements, recommendations), switching it to shade-only for a week, and comparing primary-task interrupts against downstream actions. If actions hold and interrupts fall, the old channel was buying empty opens.

Related

• Within the group: H5.02.1 Interruption cost rises with how deep the task has gone · H5.02.2 Subtask boundaries are cheaper moments to interrupt
• Adjacent: H5.01 Urgency grading · H5.07 Push frequency · H5.06 Actionable notifications
• Search terms: interrupt vs inform · silent delivery · low-value notification