Molting — Fashion Tech
Cutting onboarding drop-off from 56% to 39% with an avatar-first flow
Service
Molting (Avatar-based virtual fitting platform)
Feature
Avatar Onboarding
Role
Role
Role
UX Strategy, Product Thinking, Interaction UI
Worked with
Worked with
Worked with
Product managers, Developers (Android, iOS)
Summary
Summary
Onboarding was blocking the core product
Onboarding was blocking the core product
The Challenge
56.2% of users dropped off before reaching the styling phase. Analytics revealed that excessive input friction prevented users from discovering the product’s core value.
56.2% of users dropped off before reaching the styling phase. Analytics revealed that excessive input friction prevented users from discovering the product’s core value.
56.2% of users dropped off before reaching the styling phase. Analytics revealed that excessive input friction prevented users from discovering the product’s core value.
The Solution
Restructured the flow to lead with a visual result. Instead of starting with a blank form, users now begin by fine-tuning a pre-set avatar, which keeps them engaged and reduces the perceived effort of the process.
Restructured the flow to lead with a visual result. Instead of starting with a blank form, users now begin by fine-tuning a pre-set avatar, which keeps them engaged and reduces the perceived effort of the process.
Restructured the flow to lead with a visual result. Instead of starting with a blank form, users now begin by fine-tuning a pre-set avatar, which keeps them engaged and reduces the perceived effort of the process.
Key Metrics
56% → 39%
Drop-off
56% → 39% Drop-off
56% → 39%
Drop-off
Lowered the drop-off rate by showing a pre-set avatar draft right from the start.
+18.5%
Styling attempts
+18.5% Styling attempts
+18.5%
Styling attempts
Earlier value delivery drove higher engagement in the virtual fitting feature.
Current avatar creation flow (As is)
Current avatar creation flow (As is)
Problem
Problem
Problem
Critical drop-off at entry phase
Critical drop-off at entry phase
The Barrier
56.2% of users dropped off before reaching the styling features.
56.2% of users dropped off before reaching the styling features.
The Barrier
56.2% of users dropped off before reaching the styling features.
Friction Points
30% of users dropped at body input; another 18% at body shape.
30% of users dropped at body input; another 18% at body shape.
30% of users dropped at body input; another 18% at body shape.
The Real Question
Why did users disengage before seeing their avatar?
Why did users disengage before seeing their avatar?
Why did users disengage before seeing their avatar?
Insight
Prioritizing reward before effort
Prioritizing reward before effort
Prioritizing reward before effort
Hypothesis vs. Reality
We initially assumed the flow was too long. So we prototyped a shorter version. In usability tests, users still stalled at the same input steps.
We initially assumed the flow was too long. So we prototyped a shorter version. In usability tests, users still stalled at the same input steps.
We initially assumed the flow was too long. So we prototyped a shorter version. In usability tests, users still stalled at the same input steps.
Hypothesis A
The input burden is too high early on
1
2
3
4
5
6
Hypothesis B
Body shape selection causes hesitation
1
2
3
4
5
6
The turning point
Both hypotheses were partly right, but the root cause was deeper: users lacked the motivation to provide data without an immediate visual result.
Both hypotheses were partly right, but the root cause was deeper: users lacked the motivation to provide data without an immediate visual result.
Both hypotheses were partly right, but the root cause was deeper: users lacked the motivation to provide data without an immediate visual result.
What Usability Testing Revealed
What Usability Testing Revealed
What Usability Testing Revealed
1
2
3
4
5
6
-30% Drop-off
-18% Drop-off
Basic Input
Longer dwell time and erratic tapping on the first screen.
Longer dwell time and erratic tapping on the first screen.
Unclear Value
Users feel "cost" before seeing any "gain."
Users feel "cost" before seeing any "gain."
Body Input
Frequent micro-adjustments of sliders followed by exit.
Frequent micro-adjustments of sliders followed by exit.
Resistance to "Blind" data.
Reluctance to share info without feedback.
Reluctance to share info without feedback.
Body Shape
Rapid toggling between icons without a final choice.
Rapid toggling between icons without a final choice.
Decision Paralysis
Difficult to choose without a reference point.
Difficult to choose without a reference point.
Basic Input
Longer dwell time and erratic tapping on the first screen.
Unclear Value
Users feel "cost" before seeing any "gain."
Body Input
Frequent micro-adjustments of sliders followed by exit.
Resistance to "Blind" data.
Reluctance to share info without feedback.
Body Shape
Rapid toggling between icons without a final choice.
Decision Paralysis
Difficult to choose without a reference point.
n=8
n=8
To validate this, we tested a prototype with the avatar shown first: all 8 participants completed the remaining inputs without stalling.
To validate this, we tested a prototype with the avatar shown first: all 8 participants completed the remaining inputs without stalling.
To validate this, we tested a prototype with the avatar shown first: all 8 participants completed the remaining inputs without stalling.
Solution
Prioritizing visual results over data entry
Prioritizing visual results over data entry
Prioritizing visual results over data entry
Action 1
Action 1
Reward-First Onboarding
Reward-First Onboarding
Moved the face scan to the start to deliver an immediate Avatar Draft.
Starting with a visible draft gave users a reason to finish the remaining steps.
Moved the face scan to the start to deliver an immediate Avatar Draft. Starting with a visible draft gave users a reason to finish the remaining steps.
Moved the face scan to the start to deliver an immediate Avatar Draft.
Starting with a visible draft gave users a reason to finish the remaining steps.
Before
Before
Before
Users had to complete multiple input steps before seeing their avatar.
After
After
After
Users can see and start adjusting their avatar right from the first step.
Action 2
Action 2
Inputs as refinements
Inputs as refinements
Repositioned the detailed body and shape inputs to follow the initial avatar display. Instead of defining data from scratch, users now simply adjust the visible avatar, making the process feel much lighter and more intuitive.
Repositioned the detailed body and shape inputs to follow the initial avatar display. Instead of defining data from scratch, users now simply adjust the visible avatar, making the process feel much lighter and more intuitive.
Repositioned the detailed body and shape inputs to follow the initial avatar display. Instead of defining data from scratch, users now simply adjust the visible avatar, making the process feel much lighter and more intuitive.
Action 3
Action 3
Immediate feedback loop
Immediate feedback loop
Every adjustment now triggers real-time visual changes on the avatar. This immediate response allows users to see the direct impact of their inputs, keeping the process engaging until they reach the styling stage.
Every adjustment now triggers real-time visual changes on the avatar. This immediate response allows users to see the direct impact of their inputs, keeping the process engaging until they reach the styling stage.
Every adjustment now triggers real-time visual changes on the avatar. This immediate response allows users to see the direct impact of their inputs, keeping the process engaging until they reach the styling stage.
Impact
Drop-off cut from 56% to 39%
Drop-off cut from 56% to 39%
Drop-off cut from 56% to 39%
*Measured for new users over 8 weeks after release
*Measured for new users over 8 weeks after release
Shift in user behavior
Users began treating the process as a personalization task rather than a form-filling chore
Users began treating the process as a personalization task rather than a form-filling chore
Shift in user behavior
Users began treating the process as a personalization task rather than a form-filling chore
Lowered entry barrier
The decrease in drop-offs indicates that an upfront visual result justifies the effort of data input for users.
The decrease in drop-offs indicates that an upfront visual result justifies the effort of data input for users.
The decrease in drop-offs indicates that an upfront visual result justifies the effort of data input for users.
Impact on core activation
The 18.5% increase in styling attempts suggests that fewer users were lost before reaching the core feature.
The 18.5% increase in styling attempts suggests that fewer users were lost before reaching the core feature.
The 18.5% increase in styling attempts suggests that fewer users were lost before reaching the core feature.
Reflection
Reflection
Reflection
Prioritizing immediate value in UX
Prioritizing immediate value in UX
Prioritizing immediate value in UX
Design for momentum
I learned that reducing friction isn't just about cutting steps; it's about building momentum. By showing a visual result upfront, I gave users a reason to stay and complete the flow.
This taught me that giving value early is often more effective than simply making a process shorter.
Scalability of reward-first logic
The 'Reward-first' approach has potential beyond onboarding. I plan to apply this logic to other complex features, such as the styling feed. By showing immediate style recommendations before asking for detailed preferences, I can lower the entry barrier across the entire product.
The 'Reward-first' approach has potential beyond onboarding. I plan to apply this logic to other complex features, such as the styling feed. By showing immediate style recommendations before asking for detailed preferences, I can lower the entry barrier across the entire product.
Balancing automation and personalization
While presets were effective in reducing drop-offs, the next challenge is finding the right balance between automated ease and user agency. My goal for the next iteration is to ensure users feel enough creative control to stay connected to their avatars for the long term.
While presets were effective in reducing drop-offs, the next challenge is finding the right balance between automated ease and user agency. My goal for the next iteration is to ensure users feel enough creative control to stay connected to their avatars for the long term.