Breaking the Ice with AI
A feature designed to reduce ghosting, boost message rates, and improve early retention.
My role
Staff Product Designer
Team
5 people (PM, Eng, Content, UXR)
Date
2023
Duration
3 weeks
Responsabilities
UX, Research, Interaction Design
Context
BFF is a standalone app within the Bumble apps ecosystem designed to help people make new friends, not dates. But starting a conversation with a stranger can feel awkward, even when you've matched based on shared interests. On BFF, many matches were ending in silence, particularly in the first 24 hours. This early silence often led to ghosting and drop-off, especially among women — a critical retention segment. We needed a way to help users break the ice in a natural, low-pressure way and move from match to message quickly.
Problem & Opportunity
Ghosting and stalled chats were a top complaint from BFF users. Many felt unsure of what to say, or didn’t feel confident sending the first message. At the same time, we knew that when a user sent a message within 24 hours of matching, they were significantly more likely to stay. This opened a clear opportunity: give users a smart nudge to help them send a first message based on what they already have in common.
Insights
- Matches weren’t converting to chats
- Most ghosting occurred in the first 24–48 hours
- Users with shared interests or life stage tags were more likely to engage
- Icebreakers were one of the most requested features in past research
Matches aren’t turning into conversations, users don’t know how to start, and the app doesn’t offer the right support at the right moment.
Research & Strategy
We started by reviewing UXR studies, in-app feedback, and retention data across the match-to-chat funnel. A key insight emerged: users often matched with intent but stalled at the first message, unsure of what to say. To dig deeper, we ran diary studies and interviews with new BFF users, uncovering a strong preference for contextual, mutual prompts over free-text or generic icebreakers. Working with UXR and our Content Designer, we created a content safety matrix to determine which profile tags felt safe and inclusive to generate questions from excluding more sensitive topics like “fertility journey” or “therapy.” To identify the right delivery moment, we analyzed first-session funnel data, session recordings, and heatmaps. This revealed a consistent drop-off after matching, but before users returned to swiping, the ideal window to introduce a lightweight nudge.
Solution
We introduced AI-powered Icebreakers, contextual prompts generated based on users’ shared profile tags like interests, life stages, and location. If a match hadn’t started a conversation, we prompted one of the users to send a ready-made question. Both users could view the question, but answers only became visible once both had replied, creating a mutual, low-pressure entry point. The prompts were generated using the OpenAI API, customized with our tone of voice, safety filters, and tag-matching logic. This allowed us to scale human-like, relevant questions while ensuring they felt warm, respectful, and on-brand.
Monetization Consideration
While designing this feature, I also explored how it could support Bumble’s broader monetization strategy without compromising user experience or scalability.
I initially proposed two models:
– A daily limit on AI-generated questions, with the option to unlock more via Bumble Premium
– A one-time purchase model, offering 10 extra questions for $5
Ultimately, we introduced the feature as part of the Premium subscription, giving paying users access to unlimited prompts — while offering a limited number of daily questions to free users. Given Bumble’s large user base, I also worked closely with the product and engineering teams to limit the number of OpenAI API calls per user. This helped us maintain cost control while ensuring the feature could scale responsibly and sustainably across markets.
Results
Icebreakers helped improve performance across early engagement metrics:
- Matches that converted to chats: +4%
- Women sending a message within 24h: +6%
- Retention uplift (D7) for users who used the feature: +3.5%
- Increase in Premium subscriptions +7%
It also quickly became one of the top-referenced features in App Store reviews and user feedback.
Next Project
Player redesign
A new player to enhance the music experience and give users more control over the lyrics.
Send message© Frank Rapacciuolo 2022 Product designer