Repeating the tour cannot fix a missing entry

What it is

A hidden entry, a wrong name, or an entry absent at the relevant moment is a structural defect. Playing the tour, bubble, or banner again for the same people may bump adoption during the campaign week, then it falls back. Repeated guidance cannot patch an entry: teaching is an episode; the entry is the environment. If the environment stays, the episode must be re-staged every time, and re-staging is quickly coded as noise. This entry is the wrong remedy after discovery fails. It is not what happens when hint volume crosses a threshold, and not checklists turning into burden.

Why it happens

Teaching can write the entry’s coordinates into working memory for the seconds it is on screen. If those coordinates never become stable scent in the UI, the next task still searches the environment and finds the same hiding place. Repeating the playback tries to fake memory with spaced review, but what is reviewed is “someone once pointed,” not “I see the button.” Worse, repetition lowers the credibility of all teaching: people immunize against overlays to dodge the patch, and a later genuinely timely sentence is collateral. Adoption spikes in a campaign week are often one prompted try, not a repaired discovery—a week later the unchanged entry goes quiet again. Running discovery as a growth campaign lets a team treat teaching as the lever and leave the entry for next quarter.

Where it stops holding

When the entry is already present and only the aftermath is invisible (a wait for permission or data), repeating copy addresses expectation, not discovery; that belongs as one persistent sentence by the entry, not as campaign replay. Legally required restatement of risk before every action is notice, not a discovery patch. Briefly using teaching as a probe in an experiment (will people use it once pointed) is fine, but the probe must expire; overtime it should be read as “they still will not take this path” or “they still cannot find it,” not turned into a standing channel.

Applying it

• When adoption is low, freeze new automatic teaching and inspect whether the entry is on the current task screen, whether the label matches the user’s verb, and whether it is present when relevant.
• Replace a planned “another round of new-feature bubbles” with one entry change; teaching at most a before/after probe that expires.
• Do not take “opens this week” as evidence discovery is fixed; watch conditional adoption after teaching is removed.
• Verify under the same demand signal: compare “teaching only” with “entry change only” two weeks later. The teaching-only line falls back; when it does, do not add a third round.

Related

• Within the group: H2.05.1 Discovery depends on a visible entry, not a lesson · H2.05.2 Usage data locates features that were never found
• Adjacent: H2.07 Hint Restraint · H2.03 Progressive onboarding · G2.08 Predictability of navigation items
• Search terms: discoverability patch · tour cannot fix IA · feature coaching