v1 · The Personas
Some people come for the garage. Some to organize the whole house. Some because they lose things constantly — the ADHD brain that needs an external memory. One early question can serve all three, if each answer visibly changes what comes next.
Why ask at all
The strongest evidence is the Headspace × Irrational Labs controlled test (8 months, 5 arms). Asking a few questions before an identical recommendation lifted course-starts from 31% → 63% overall — about +7.6 points from the "feeling seen" effect alone, zero real personalization.
Translation: the question earns the first action. Delivering a reflected-back payoff after the first snap ("Filed to your garage — now ask me where anything is") is what earns the second session.
The design
Single-select (you can only branch to one flow) · 4 options + an escape hatch · framed as goal/trigger, never demographics. The "limit to 7±2" rule doesn't apply here — a visible list is recognition, not memory recall. Cap at four because each must earn a distinct downstream, not because of a mis-cited law.
Skip / "Just exploring" routes to a good generic whole-home first run — never a dead end. Let users change it later in Settings; needs evolve.
The non-negotiable
Otherwise it's fake personalization and users feel it. Even if v1 can only differ copy + sample query per branch — ship that; it still buys the perceived-fit lift. Deepen as you instrument.
The wedge
This is a large, self-identifying community (ADHD TikTok/Reddit) and arguably NoonFox's sharpest positioning: the app is literally a software version of the workarounds this brain already invents — clear bins, drop zones, "make the environment do the remembering."
Do
Don't
Honesty note: the causal copy claims are untested for NoonFox specifically — treat the ADHD branch as hypotheses to A/B, with speed-to-first-find and D1 return as the metrics that matter most for this segment. Tone target: Tiimo's warm "external brain," not Inflow's clinical register.
Prior art
One motivation question, ~6 options (Travel, Career, School…), single-select. Feeds notification personalization. Sign-in pushed to the end. Runs ~1,200 experiments/yr.
~20-step quiz → an affirming results screen (BMI + projected graph) that reflects answers back. The payoff justifies the questions. Sign-in last.
A weight-prediction chart that updates as you answer makes input feel consequential. The updating artifact is the mechanism — not the length.
Deliberately few questions (2–3), small option sets. Proof that short + well-branched beats long-for-its-own-sake.
The shared pattern: one motivation question, single-select, small option set, and an immediate reflected-back payoff. NoonFox's payoff is the first successful natural-language find.
Receipts