Zero-result states must explain why, not only report the count

What it is

“Found 0” only states a count. Zero results must explain why translates that count into a cause the user can act on: the terms did not match in this collection, a filter sliced the set empty, the current scope is too narrow, or permission hides the hits. Without a cause, 0 is read as “this thing does not exist,” and the next move is noise—changing an unrelated word, or leaving. Hearst treats empty-result pages as having to explain what happened, not merely report absence.

The explanation addresses an empty set that has already happened. It is not blocking empty facet combinations before a click, and it is not previewing a count before a filter is applied.

Why it happens

An empty set cuts the feedback loop of query → results → reformulation. People diagnose the last shot from the results; with zero rows the only diagnostic material is a count. A count does not point: wrong word, wrong scope, wrong filter, object absent—four repairs that are nothing alike. Working memory picks the cheapest story, usually “nothing,” and finding stops. Writing the cause as a list of current constraints (query, applied filters, scope) reconnects diagnosis—the user can ask which constraint is responsible for this 0.

The cause also has to be falsifiable. “No matches” is the count restated. “发票作废 does not appear in titles or bodies in this library; the filter This year is on” names two removable culprits. Vague comfort (“try other keywords”) looks like a cause and actually returns the cost of generating the next query.

Studying it

Measure which class people attribute 0 to, and whether that attribution matches the real cause.

• Paradigms: the same empty set with different copy (count only / list current constraints / wrongly imply the object does not exist); in query logs, the next action after zero (reword, drop a filter, leave) against the true cause of emptiness. Hearst on empty-result interfaces.
• Independent variables: whether the responsible constraint is named, whether “terms missed” is distinguished from “filters emptied the set,” whether missing permission is phrased as “no results.”
• Dependent variables: attribution accuracy, rate of wrongly concluding the object does not exist, whether the next action aims at the true cause.
• Methodological note: if the lab sheet promises “the target is definitely here,” people will not choose “does not exist,” and the attribution distribution is warped. Mix empties that are truly absent with empties caused by filters. Score “understood the cause” separately from “willing to continue”: some people understand and still leave, which is a recovery-action problem.

Where it stops holding

When the engine itself does not know why it is empty (timeout, broken index), do not invent a query-side cause; say the system failed, or people will edit an innocent word. When the object truly is absent, the cause is absence—do not hang it on a filter. When several causes stack (terms off and filters tight), list them all; reporting only one steers repair to the wrong cut. Granularity stops at constraints the user can remove; do not narrate inverted indexes or tokenizers.

Applying it

• On a zero-result page, name in a short sentence which of the current query, applied filters, and scope jointly produced the empty set; do not only write “no results.”
• Use different sentences for “terms missed” and “a filter emptied the set”; the latter should make visible which chip is responsible.
• Do not phrase permission-hidden hits as “no matches,” or people will believe the object does not exist.
• Verify by creating three empties—wrong word, over-narrow filter, no permission. If the copy is identical for all three, the cause was not explained. Ask someone uninvolved “what would you change next”; that item should match the true cause.

Related

• Within the group: G3.07.2 Zero results should offer concrete actions that relax constraints · G3.07.3 A zero-result page is a guide, not a dead end
• Adjacent: G1.06 Faceted classification · T2.05 Empty-state and zero-result copy · G3.04 Spelling correction
• Search terms: zero results · empty search · query reformulation