Twindly

Enhancing Match Rates with AI Feature on Twindly

Twindly is an AI-powered dating app designed to give overlooked but interesting users a second chance at connection.

Timeline :

4 Months | Jan 2024 - Mar 2024

My Role :

User Research, User Flow

Wireframe, High-Fidelity,
Prototypes, 

UX & Visual design, User Testing

Product Link

Product Link

Overview

Twindly is an AI-powered dating app designed to give users who may not stand out on traditional dating platforms but have unique personalities worth exploring a second chance at connection. In its initial MVP launch, Twindly introduced an "AI twin" feature, allowing users to chat with an AI version of potential matches. However, this feature did not significantly boost match rates due to low engagement.

Our task was to identify the reasons behind this issue and develop strategies to enhance user interaction and improve match outcomes.

1

Timeline

Timeline

The project spans for 3 months. The design team works alongside founder and PM

The project spans for 3 months. The design team works alongside founder and PM

2

The Team

The Team

The product team has two designers

The product team has two designers

3

My Role

My Role

I led the team as the senior product deisgner and supervised one intern product designer

I led the team as the senior product deisgner and supervised one intern product designer

The Outcome

After the redesign, we did one round of thorough user test with a group of users (16), and get the positive feedback from users. Twindly is currently still in developing phase after our design team has delivered the final product design to the engineering team and is expected to launch in the first quarter of 2025.

Problem Statement

for previous version

On user profile swiping page, user can choose to talk to the AI twin to know more about that person.

Low User Engagement with AI Twin

Users lacked motivation to interact with their AI twin, leading to low usage of the feature and little effect on match rates.

Limited Data for AI Optimization

Limited user input made it hard for the AI twin to create engaging conversations, reducing user interest and matches.

Design Questions

How can we boost engagement with the AI twin to increase match rates and get enough data on training better AI twin on Twindly?

How can we boost engagement with the AI twin to increase match rates and get enough data on training better AI twin on Twindly?

Our Solution

in a nutshell

Instant Match via AI Twin Chat

This serves as a big motivation for users to engage in chatting with AI twins so they can matching with the user without sending a request. (Our user research of 20 users also indicates that both male and female users are prone to this feature and don't feel disturbed by the instant match, as long as the user has put efforts in talking to their AI twin at first)

AI Twin Built During Onboarding

User must engage in creating an AI twin during onborading and they can set the number of messages that others must exchange with my AI Twin before they can instantly match with me.

Design Process

Process Overview
+ Timeline

The process follows the design thinking methodology, starting with problem framing and progressing through design and evaluation stages.

01
User Research

Based on user interviews and user journey map, we identified the following two primary user groups:

02
User Flow

The previous homepage was a user profile swipe page like most dating app and the user has the option to send like or dislike with the option to talk to the AI twin.

We had two proposals:

Proposal A

User can send match request after talking to the AI twin and the other user can review their chat to choose to match or not match with them

The updated homepage features a user pool, allowing users to choose someone they want to invest time in by interacting with their AI twin.

The match review page consolidates match request information into a card, allowing users to expand it to view detailed interactions from the requester and their AI twin.

Proposal B

User can send direct match with the target user once they have talked with their AI twin for a certain amount. (We adopted this proposal after conducting user testing with 7 users (4 female and 3 male).

03
Wireframing

Optimizing itinerary generation experience to boost conversion rates and enhance business collaborations

04
Key Features

Key features for out final product delivery

1

Send Match Request

Send Match Request

Boost your match success by starting a conversation with a user's digital twin. Discover more about the person who caught your eye through a fun, interactive experience.

Boost your match success by starting a conversation with a user's digital twin. Discover more about the person who caught your eye through a fun, interactive experience.

2

Build My Twindly

Build My Twindly

Easily create your own digital twin by chatting with it as if it were a real person. Let others get to know you on a deeper level without the need to engage in every conversation.

Easily create your own digital twin by chatting with it as if it were a real person. Let others get to know you on a deeper level without the need to engage in every conversation.

05
User Feedback

Match Rate (Passion Pulse) Position

"I didn’t immediately notice the match rate; it would be helpful if it were more prominent."

Profile Layout as a Card

"It’s hard to get an overview of someone’s profile without scrolling a lot."

Tag Scrolling with Right Arrow

"It’s a bit inconvenient that I can’t scroll through tags easily."

Tag Number Indicator and Match Requests

"I wasn’t aware of how many requests I had pending. A number indicator would be helpful."

AI Icon for Messages Sent by AI

"It’s a bit confusing when I see messages from the AI without any clear distinction."

06
Iterations From Previous Version

We collect user feedback from our testing sessions with over 15 participants and make updates based on their input.

1

User profile layout

2

My Twindly page layout

3

Chat list layout

07
Final UI

We collect user feedback from our testing sessions with over 15 participants and make updates based on their input.

07
Reflection & Take Away

1. The Power of Iterative Design

  • Reflection: The process of testing and refining helped us improve features based on real user feedback, showing how important it is to stay flexible and make changes along the way.

  • Takeaway: Iterative design not only fine-tunes features but also deepens our understanding of what users want. In future projects, it’s essential to allow time for multiple testing rounds to create a smoother, more user-friendly experience.

2. Balancing New Ideas with Familiar Patterns

  • Reflection: Adding innovative features like the AI digital twin and Passion Pulse was exciting, but we needed to make sure they felt easy and intuitive for users.

  • Takeaway: When bringing in new concepts, balancing them with familiar design patterns helps users feel more comfortable. For future projects, making innovative features feel approachable is key to helping users adopt them quickly.

4. The Importance of Clear Visual Hierarchy

  • Reflection: Placing key information, like the match rate, in a prominent position helped users find what mattered most, highlighting the need for a clear visual order.

  • Takeaway: A strong visual hierarchy allows users to access essential information quickly. In future designs, prioritizing the visibility of key information will improve usability and user satisfaction.

About

Thank you so much for your valuable time! Please feel free to reach out!

About

Thank you so much for your valuable time! Please feel free to reach out!

About

Thank you so much for your valuable time! Please feel free to reach out!

LU Jiawei

& more

LU Jiawei

& more