Record that a cue was seen and do not replay it

What it is

Once a progressive intro has fired at first relevance, the system must record seen and must not auto-play that same cue for that person again. Seen includes read, skipped, and dismissed—if the teaching layer was shown, the speaking slot is spent. This entry is the state machine: the intro is a one-shot event, not a replay every time the situation returns. It is not “the person can now do it”—that takes later unassisted success. It is also not the restraint rule that stops hinting mastered features; that rule watches use. This one only watches whether the lesson was already put on screen.

Why it happens

Relevant situations recur. Without a seen flag, just-in-time teaching degrades into a shout on every pass, and people code the skin as ignorable. The flag makes “first” an executable past tense: speak on first relevance, stay silent on the second. Skip must write seen too. Skip means this refusal spent the slot; if skip does not count, the next relevance comes back, and refusal becomes an endless job. The record must also cross sessions and surfaces: an intro from a list must not return in a different skin on a detail view. Losing the flag on a new device or a cleared cache makes returning users pay attention they already paid, which hurts more than a missed intro—missed teaching can still be found at the entry; repeats train “always dismiss this layer.” Seen must not mean “never offer the explanation again”: after the automatic layer stops, summonable help by the entry remains; it just stops interrupting.

Where it stops holding

If the interaction itself changed incompatibly (entry moved, consequences changed), an old seen must not block a new necessary notice—but that notice is a new item, not the old lesson replayed as an “update.” Shared devices or demo accounts that share one seen hide all intros from later real users; the flag should follow an identifiable user, with demo mode reset separately. Legal or safety copy may still need to appear before every high-consequence action after it has been “seen”; that is not progressive teaching and does not use this one-shot flag. A layer that was on screen for 200 ms before the system dismissed it should not count as seen, or a hitch will swallow the only speaking slot.

Applying it

• Give each automatic intro its own seen key; write it as soon as the layer is successfully shown (including skip and close); sync it with the account, not with one install’s memory.
• Share that key across surfaces of the same feature; do not keep per-page flags that replay a reskin.
• After the automatic layer stops, keep the same explanation summonable at the feature’s entry, now on request.
• Verify: after a person has seen or skipped once, return to the same relevant situation in a later session. If the automatic layer returns, the flag was not written, not synced, or split by page. Also check replay after a device switch.

Related

• Within the group: H2.03.1 Introduce a feature the first time it becomes relevant · H2.03.2 Progressive coaching needs a product-wide frequency budget
• Adjacent: H2.04 Skippable and replayable guidance · H2.07 Hint Restraint · E4.14 Coach marks
• Search terms: seen flag · do not replay · onboarding persistence