top of page
Company

Branch Metrics

My Role

Senior Product Manager

Key Contribution

Product Specification, Product Roadmap, Functionality Requirement, User Stories, Prototype, Beta Releasing

Project Duration

2023/05 - 2024/09

Branch operates in two key areas: supporting Android in-built launchers by delivering personalized content and intelligent mobile assist, and providing a mobile attribution and deep linking service. It helps businesses create seamless, cross-channel user experiences and powers over 100,000 apps globally.

unnamed.png
What is Branch Metrics?

Content Serving Platform

Mobile Measurement Partner

  • Ad serving platform on Android mobile launchers

  • Android mobile intelligent assist

  • Mobile attribution analyses and insights

  • Mobile linking platform 

Screenshot 2024-09-23 at 12.41.38.png

Branch.io

Customer Problems

Annoyed by random content recommendations

Branch was offering organic content based on recency and frequency, and ad content randomly, which caused users to become annoyed by receiving irrelevant and non-personalized content recommendations.

1~.png

Can't receive personalized content at the right time

Even though we offer personalized content, users still don't actively engage because it doesn't meet their real-time demands. Imagine if you received a message update from a dating app during work hours... 😅

2~.png

Lack motivation to engage more with global search

The Android global search DAU was only around 3%, which was extremely low because users lacked additional motivation to engage with services that could make their daily lives easier.

3~.png
Product Vision

  • Provide highly personalized content to the RIGHT user at the RIGHT time in an adaptive creative format.

  • Leverage AI to encourage users to engage more with global search on Android.

Core Experiences

Dynamic Data: Leverage the user's demographic information, device data, and in-app behaviors to predict personalized organic and ad content in real-time, with flexible auto-generated creatives.

Example Scenario: Imagine you're usually running in a nearby stadium on weekend mornings, using Strava for workout tracking and listening to your favorite artist's songs.

Frame 2.png

Collect Demographic Info:

Gender, Time of the day, Day of the week, Location, Wi-Fi connection, etc.

Frame 7.png

Collect On-Device and In-App Data:

App installed, App opens, Your workout records on Strava, Your favorite song type, artist, and podcast on Spotify, etc.

Frame 17.png

Direct the user to in-app content

ezgif-1-d76538bedf.gif

Personalized Content Offering:

Based on the collected and preprocessed real-time data, we use the well-trained ML model, to predict the best personalized content with flexible creative rendering and direct the user to the in-app content.

Android Assist: Leverage AI-driven assistive tools to solve more Android users' daily problems and questions with he voice entry, and encourage them to engage with Android's global search more.

Example Scenario: Imagine you want to quickly get answers to basic daily life questions, such as today's weather, app usage insights like monthly online purchases, and additional advertising-related services, such as a Starbucks coupon when you want to buy coffee.

Support Daily Life Questions

Branch Android Assist can support the common daily life questions, such as weather of the day, real-time traffic on the road, or next meeting on the calendar, which combine with Global Search interferce with responsive UI.

Provide App Usage Deep Insight

Besides common life questions, Branch Android Assist can also provide deep app usage insights, such as your monthly purchase history across online platforms, and recommend further based on your interest!

Provide Additional Benefits

Branch also collaborates with many brands across multiple verticals to provide you with additional benefits.

For example, when you want to purchase coffee and ask Android Assistant for help, besides directing you to the product page, we also offer discounts or coupons to help you get a lower price.

ML/ LLM Sytem Design & Iteration Process

Blueprints_png2x.png

STEP 1

Design ML/DL models based on business requirements and technical scope, with the models continuously trained using historical user data to predict user interests or generate new content.

Subscriptions_png2x.png

STEP 3

Collect users’ ongoing behaviors and feed the data back into the model, enhancing personalization in future recommendations or improving the intelligence of content generation.

Track and collect users' data, including user behaviors, purchase history, browsing patterns, and more. Then preprocess the data through cleaning and normalization.

STEP 2

CPQ_No_Products_png2x.png

The recommendation or response systems fetch real-time result data, perform creative rendering, and finally present it to the user.

STEP 4

Security_Center_png2x.png

Real-Time Attribution

Besides predictions based on these behavior models, we also use real-time attribution to analyze data transferred from user behaviors, such as clicks or non-clicks. This data is recorded and fed back into the models, enhancing personalization in future recommendations and content generation.

Target Segmentation - User funnel

Screenshot 2024-09-23 at 23.52.34.png

Unique Approaches

Leverage users' in-app deep data, which reflects their interests and demands, combined with ML & DL models to deliver more personalized and responsive content.

Frame 54.png

Enhance personalization while improving advertising effectiveness, balancing both the better user experience and maximized business and revenue growth,

Frame 55.png

Facing the deprecation of device IDs, the system is also flexible for the on-device solution without reliance on IDFA or GAID, while ensuring compliance with policy requirements.

Frame 56.png

Concern & Risk

Risk.png

Legal concerns for advertisers and user permissions

The user data we collect requires permission, especially MMP data.

Update the Ad Manager terms and conditions agreement to ensure effective consents from onboarded customers.

High

High

Challenges for offering incomplete insights

Comprehensive content generation requires holistic app coverage.

Try to cover as many verticals as possible across apps in the closed beta as MVPs to gather comprehensive insights.

High

Mild

Deprecation of device IDs for ad content tracking

The deprecation of device IDs will significantly affect ad content tracking.

Discover on-device solutions, such as SDK-to-SDK integrations, that do not rely on device IDs.

High

High

What Did I Contribute?

  • Develop the product solution with a clear business scope: Lead the team to build the Dynamic Data and Android Assist features, which I proposed, and developed from scratch.

  • Coordinate cross-functional teams:

    • Marketing team - Collaborate on the closed beta program, including customer onboarding and follow-up.

    • ​​

    • ​

    • ​

    • ​

    • ​​

    • ​

    • ​​Engineering team - Participate in model design, functionality requirement setup, and model iteration.

    • Data Science team - Jointly analyze the model and conduct use case testing to ensure product quality.

    • Enablement team - Collaborate on internal training to ensure the field team can support customers timely.

  • Significantly enhance the platform's content personalization and AI-driven intelligence support.

Screenshot 2024-09-24 at 01.44.17.png

Product Launch Result

Dynamic Data - Personalized Content

Android Assist - Intelligent Mobile Support

  • User acquisition:

    • Increased CTI 19%.

    • Drove 200,000 incremental installs for Spotify, Reddit, Expedia in India within 1 month.

  • ​Re-engagement:
    Compare to other Branch ad types

    • Increased CTR 21%,

    • Increased CR 16%,

    • Increased ARPU 12%.

  • Increased DAU for Android Global Search from 3% to 9.7%. ​

  • Increased ad purchase conversions by 6.4%.

bottom of page