Partial batch failure needs per-item status

What it is

Twenty items submitted, seventeen through, three not: the result cannot collapse into one lamp. Per-item marking means each item still shows success or failure, and a failed item still shows its own cause. People need that list to decide which rows to touch next. This entry is only about how status is listed. It is not about retrying failed items alone, and not about whether partial success may be labeled as a failed batch.

Why it happens

A batch folds attention from objects into a set. Failure happens at the object: permission, conflict, remote refusal, each different. One set-level color flattens those differences; working memory keeps only “it didn’t work,” so people either redo the batch or stop checking. Per-item status splits the set back into objects; a glance separates the two classes, and causes need not be guessed. The list also supplies a spatial anchor: failed items stay in place or can be sorted together, rather than vanishing into a banner.

Studying it

On the same mixed result, compare one summary, two counts (ok / failed), and per-item status.

Independent variables: grain of status, whether the cause appears on that item, whether the list can be filtered by status. Dependent variables: accuracy at pointing to failed items, times a successful item is redone, time to finish checking.

If lab items have no recognizable names, per-item status cannot measure discovery. Use objects the participant just selected.

Where it stops holding

Huge N cannot be drawn on one screen; offer a failure filter and an export rather than falling back to one total. While animation or progress is still running, unfinished items are in-progress, not failed; keep three classes. When some items are still unknown, mark unknown; do not paint them failed and then flip to success.

Applying it

• On the batch-complete page or the original list, each item shows success / failure / in-progress, and a failed item carries its own cause sentence.
• Offer a “failures only” filter whose count matches the list.
• Do not let a banner replace the list; a banner may summarize, details live on the item.
• Verify with a deliberate mixed result, hide the summary, and ask “which rows did not go through.” If they point wrong or can only repeat a number, per-item status did not land.

Related

• Within the group: H3.13.2 Failed items need their own retry, not a full redo · H3.13.3 Partial success must not be labeled as total failure · H3.13.4 Say whether the failed items already had side effects
• Adjacent: H8.02 Batch operations · H1.16 Post-submit result presentation · H3.02 Three elements of error messages
• Search terms: partial failure · per-item status · batch result