Retention.
End-to-end ownership · I led the retention lifecycle for Meta AI glasses — and shipped the 30-day recap that proved the system.
End-to-end ownership · I led the retention lifecycle for Meta AI glasses — and shipped the 30-day recap that proved the system.
I mapped where users drift, then chose the right move at each point — what to surface, to whom, and when. I shipped a series of interventions across this journey; the day-30 recap below is the one I'll show in depth.
Meanwhile experiments were fighting each other — everyone reached for the full-screen bottom sheet. So I defined the operating logic: match a user's maturity and a feature's value — scored by a Retention Rating — to the right component intensity, so the next nudge was always the next best thing untried.
Of everything I shipped across the lifecycle, this is the one I'm proudest to show — a Wrapped-style story for your glasses. Five pages recap your month, then point you at what to try next.
Not everything I designed made the final recap. Some calls were mine, some came from Legal and privacy review — but knowing what to leave out is the judgment, not the loss.
I ran all four phases — concept, prototype, user testing, production — with AI in the loop. AI expanded directions and pre-tested the flow; I built the front-end in Claude Code and committed to the same codebase as engineering, then fixed UI directly in QA instead of filing tickets.
AI pushed execution toward zero. Judgment became the bottleneck — choosing the direction, holding the context, owning the quality.
As an experiment, the entry showed once, then stepped aside for other experiments. Users still dug it out of their notification history — some, again and again.
A pushed promo had quietly become a personal artifact — something people chose to reopen. I read it as observed behavior, not proven motive, and used the signal to reshape the retention roadmap that followed.
The right moment, the right signal, and the right level of interruption can turn retention from a push into a pull.