A mismatch between expectation and arrival lowers trust in the whole site

What it is

Click Help and land on a marketing page, and people do not only file “this button was mislabelled.” They downweight every other label on the site. One failed preview leaks to slots not yet clicked: navigation is treated as one speaker, so a falsification in one place discounts scent elsewhere. Trust here is not brand affection. It is whether labels are still worth using as wayfinding evidence.

A single wrong arrival can be a misclick. Repeated wrong arrivals become a strategy change: stop reading the nav, switch to search, or leave.

Why it happens

Scent is a conditional probability on hidden content. One arrival supplies a counterexample and the posterior drops. Because words in the same nav are treated as one speaker, the counterexample generalizes: words on this site are unreliable. Generalization saves cognition—no need to retest every slot—and spreads harm from one label to the whole row.

The drop shows up as shorter gazes on labels, more open-and-back, earlier switches to search, and hesitation even on correct labels. Fixing that one wrong label does not immediately unwind the generalization; a stretch of unfalsified use is needed before “words on this site are trustworthy” rises again.

Studying it

After one controlled wrong arrival, measure behaviour on labels that were not involved; do not only retest the item you changed.

• Paradigms: several normal nav tasks as baseline, insert one clearly mismatched arrival, then measure first clicks, open-and-back, and search switches on untouched items; afterwards ask whether the menu names are still believed.
• Independent variables: how wild the mismatch is, which trial it occurs on, whether the landing page apologizes.
• Dependent variables: hesitation and misclicks on uninvolved items, search switches, whole-site trust ratings, how many correct arrivals after a fix are needed to recover baseline.
• Methodological note: “help us test this site” in the lab makes people tolerate mismatch. Use their own goals. Retesting immediately after a fix overestimates recovery; wait a session to see whether generalization remains.

Where it stops holding

When a slot is already known to be an ad or partner entrance, the expectation model is different and a mismatch is not billed as “site labels.” Experts immune to an internal module name may not lose trust in other modules from one miss. A mobile mis-tap that can be backed out of instantly is often blamed on the finger, not the site; if back-out is impossible it is still billed to the site. New users have no baseline; the first mismatch defines “are this site’s words accurate,” and the harm is larger than for veterans.

Applying it

• Treat nav items with immediate bounce as trust incidents, not only as content problems on that page. Check the preview; if it is wrong, change the label or the membership first.
• Keep promotional jumps whose bulk is not what the label names off primary nav; promotions use an explicit campaign entrance.
• After fixing one wrong label, watch open-and-back and search-switch for the whole row for several weeks, not only clicks on the fixed item.
• Verify by staging (in a prototype) one wild mismatch, then a findability task unrelated to that slot. A rise in search escape on unrelated slots means trust already leaked past the single point. In production, monitor whole-row “back within three seconds of open.”

Related

• Within the group: G2.08.1 A nav label must accurately preview what it opens · G2.08.3 Vague labels force people to probe items one by one
• Adjacent: G1.04 Labeling systems · G2.07 Navigation consistency · G3.01 Searching versus browsing
• Search terms: broken information scent · trust in navigation · label-destination mismatch