Sarah, 34
Juggling work and childcare, Sarah needs to know instantly if a package arrived or if her toddler is near the pool.
Pain Point: Overwhelmed by generic alerts; misses important events.
Beyond Generic Alerts: Boosting Tenant Subscriptions in Offline Communities
AIGrowth & Retention
At Arlo, we noticed users were drowning in generic alerts from their security cameras. Our hypothesis? Context is king. By adding AI-powered captions (e.g., “Package delivered” or “Person at front door”), we aimed to make security feeds smarter, faster, and more valuable while driving Premium subscription growth. This case study outlines how we applied design strategy, UX thinking, and AI technology together to deliver smarter, more contextual alerts that drive engagement - without adding noise.
As the lead UX designer on Arlo’s AI Event Caption project, I worked remotely with a stellar cross-functional team to turn generic security alerts into clear, contextual insights. Partnering with Product Managers and User Researchers, we aligned business goals with user needs and validated our hypothesis through diary studies. AI/ML and Backend Engineers built the real-time captioning system, while QA ensured accuracy and reliability. Through weekly syncs and Figma collaboration, we launched smart alerts like “Package delivered” that boosted engagement and supported Premium growth, all with a simple, human-centered experience.
Project Overview
My Team
2 Mid-Level Designers, 2 Product Managers, 2 User Researchers, 2 AI/ML Engineers, 3 Backend Engineers, 1 QA Engineer.
My Role
Lead Product Designer (end-to-end ownership).
Responsibilities
- UX strategy and design execution (discovery → final handoff)
- Prototyping, usability testing, and design iteration
- Aligning features with user needs and business goals
- Optimizing for a frictionless in-app experience
Timeline
Feb - Jun 2025.
The Problem
Arlo users receive dozens of camera notifications per day—70% of these alerts lack context, leaving users confused about whether they require immediate attention. As device adoption grows and expectations for smart home security rise, users increasingly struggle to quickly assess what’s happening in their environment. This leads to fatigue, slower response times, and underutilization of subscription features.
What I Accomplished
- Designed an AI-powered video captioning system integrated into the Arlo app, summarizing key events captured by video-enabled devices.
- Integrates with notifications, feeds, and search (without exposing sensitive data).
- Led usability testing and iterated based on real user feedback to optimize discoverability, comprehension, and trust in the captions.
- Collaborated with AI/ML Engineers to ensure the captions were clear, relevant, and action-oriented.
Impact
22%
+22% Premium subscriptions in Q3 - beating our goal by 7%.
35%
35% more engagement with captioned alerts vs. standard motion alerts.
35%
NPS increased by 35% - users loved the clarity.
Discovery
It’s been a while...
We began with internal hypothesis framing and user research. Feedback from Premium subscribers showed that users often ignored event thumbnails because they “all looked the same.” They wanted faster ways to know what triggered an alert—whether it was a person, vehicle, or just a tree branch. We also analyzed customer support tickets and app analytics, which revealed: High churn among users who felt overwhelmed by too many irrelevant alerts.
Definition
Defining the North Star
Based on our findings, we defined a clear objective: Transform standard camera alerts into high-value, actionable intelligence that justifies a Premium subscription. We identified three primary goals to guide our design iterations:
- Context First: Ensure the most important information (Who/What/Where) is visible at a glance without opening the app.
- Build Trust: Design for AI accuracy by using clear, non-robotic language and providing easy ways for users to give feedback if a caption is wrong.
- Frictionless Upsell: Naturally demonstrate the value of captions to non-premium users within the existing notification flow to drive conversion.
User Research
Understanding the “Why”
We conducted 1:1 interviews with 15 Arlo users—ranging from tech-savvy early adopters to busy parents—to identify the biggest pain points in their daily security monitoring.
Mark, 45
A tech-forward homeowner who monitors multiple properties and expects precise data.
Pain Point: Wants detailed logs and quick searchability for specific types of motion.
User Pain Points
What we heard
Common themes emerged around the high cognitive load of managing home security. Users expressed that while they felt safer with cameras, the sheer volume of data was becoming unmanageable.
“I want to know it’s important before I even pick up my phone. Right now, every alert feels like a ‘cry wolf’ situation.”
“I get so many notifications that I’ve stopped checking them altogether.”
“I wish I could just see a summary of what happened instead of watching every clip.”
“It’s hard to tell which alerts are important and which ones are just the wind.”
Gain Confidence
Feature Prioritization
To gain alignment with stakeholders, I facilitated a design sprint with Product and Engineering. We mapped out possible AI use cases and ranked them by:
- Feasibility (ML readiness)
- User value (based on surveys and interviews)
- Business impact (subscription growth and retention)
Captions emerged as a clear foundational feature—enabling smarter notifications, video search, and future threat detection capabilities. We aligned on launching captions for Premium plan users first to drive growth and validate user appetite. We focused on:
- Short captions (under 5 words) for quick scanning.
- Medium captions (1–2 sentences) for more detail.
- Long captions (hidden) for future searchability.
- User consent & control (opt-in per camera).
Building Trust
We recognized that for AI-powered captions to be effective, users needed to trust the information provided. We focused on three key areas to ensure the system felt reliable and transparent:
Contextual Clarity:Provide clear, human-readable summaries that answer the ‘Who, What, Where’ immediately.
Feedback Loop:Design easy-to-use ‘Is this correct?’ prompts to allow users to refine the AI, giving them a sense of control.
Accuracy Indicators:Use subtle visual cues to indicate the AI’s confidence level, preventing ‘false alarm’ fatigue.
Explorations
I explored multiple UX directions:
- Minimalist caption overlays on thumbnails vs. text blocks beneath
- Device-level vs. location-level feed integration
- Iconography and visual cues to indicate AI confidence
- Accessibility-focused designs (e.g., screen reader compatibility and font contrast)
Through multiple design reviews and iterative testing, we converged on a solution that was:
- Non-intrusive, yet visible enough to catch attention
- Scalable to multi-device scenarios
- Compatible with Arlo’s latest app visual language and components
Ideation & Alpha Testing
We conducted alpha testing with Arlo employees and a small cohort of existing users. Every time new features were designed and implemented, we announced updates in the Slack channel to keep participants informed. After each round of feature testing, we distributed surveys to gather immediate feedback for future improvements.
Final Outcomes
Designs & Prototypes
In addition to delivering the final visual design, I created interactive prototypes for both iOS and tablet screens. These prototypes demonstrated how the interactions would look and feel, serving as a key reference for engineers to ensure accurate implementation of the design and motion details.
Education for First Time User
The final product brought together several important elements, including a guided experience for first-time users. This onboarding flow introduced the new AI event captions feature in a clear and approachable way, helping users quickly understand how it works while reducing confusion and improving the overall experience.