01 / USE CASEOOTD
The outfit poster
A direct upload moment keeps the daily action clear.
A daily social ritual shaped around authentic feedback
[SECTION 01 // OVERVIEW]
A daily outfit ritual, without the pressure to perform.
OOTD makes daily outfit sharing feel more like community than performance. I led product design end-to-end as Founding Designer and Co-Founder, partnering with engineering through an alpha-tested sprint.
End-to-end product design · Cross-functional engineering · Active alpha users
At a glance
Role
Founding Designer + Co-Founder
Scope
End-to-end product design
Validation
Active alpha users
[SECTION 02 // PROBLEM & MARKET]
Make daily fashion sharing feel safer to join.
Daily sharing can build a habit, but performance pressure and shallow likes can make posting feel risky.
COMPETITIVE TRADE-OFFS
Borrow the behavior. Rework the pressure.
Tinder’s binary swipe was rejected: fashion is subjective and context-dependent, so a yes/no verdict can reduce personal style to superficial judgment. BeReal’s urgent prompt makes the habit easy to remember, but without structured social reward it can add fatigue. OOTD keeps the daily rhythm and gives feedback more nuance.
Interaction trade-offs
Keep the habit. Change the feedback.
Posting friction
Intentional · daily
Feedback nuance
Higher · stars + context
Social anxiety
Designed to reduce
OOTD keeps a repeatable daily rhythm while using contextual tags, private feedback, and non-punitive score curves to make feedback more expressive than a like.
DESIGN APPROACH
A simple loop, designed around people.
Familiar patterns helped users get started. I worked with engineering to simplify onboarding, navigation, and first-release UI states.
From idea to daily ritual
A familiar posting loop, shaped for more thoughtful community feedback.
[SECTION 03 // ANATOMY OF THE LOOP]
Three decisions shape a healthier daily loop.
[SECTION 04 // BAYESIAN SAFEGUARDS]
A confidence anchor for early ratings.
More expressive feedback can still feel exposing.
A single 5/5 can look perfect; a single 1/5 can feel punishing. Bayesian averaging pulls low-volume posts toward the community mean, then gives the post’s own ratings more weight as votes arrive. This is designed to reduce early-score volatility—not to promise that ratings cannot hurt.
Small samples, steadier scores
See the confidence anchor at work
Keep the selected rating constant, then change how many people have rated it.
Raw average
5.00 / 5
Moves directly with every vote
Bayesian weighted
/ 5
Anchored until confidence grows
1 vote50 votes
Choose the repeated rating
W = (v / (v + m))R + (m / (v + m))C
R = raw mean · v = votes · C = community mean · m = prior strength
Illustrative settings: C = 3.5 stars, m = 10 prior votes. Calibrate with platform data before launch.
Safeguards flow
Protect confidence before a score becomes social proof.
A five-star response is more expressive than a like—but without safeguards, a single early rating can become a public penalty.
Try an early rating
01 / 2★
A low early signal
The post receives a low rating before enough context or votes exist.
02 / SAFEGUARD
Private context first
A tag-based prompt turns a blunt score into constructive, specific feedback.
03 / SAFEGUARD
Non-punitive score curve
Bayesian smoothing anchors the early score near the community mean.
04 / SAFEGUARD
Confidence grows with data
The post’s own average earns more influence only as ratings accumulate.
[SECTION 05 // REFLECTIONS & IMPACT]
What alpha feedback changed.
Alpha feedback shifted the concept from binary votes to five stars. That added nuance—and surfaced a fairness edge case: a handful of early ratings should not define a post. Bayesian smoothing became the next design direction. No retention or engagement lift was measured in this alpha.
REFLECTION / WHAT I’D CARRY FORWARD
What I’d carry into the next product.
Designing habit-forming loops without toxic social metrics.
Using Bayesian smoothing as a UX shield against performance anxiety.
Streamlining onboarding navigation under strict engineering constraints.
