Project Overview

Company: Google
Role: Lead designer
Team: 1 Designer, 2 ML researchers, 5 Engineers, 1 Project Manager
Timeframe: 3 years

Superbloom was a pioneering project I led within Google to tackle a common challenge in digital advertising: the rigidity of ad templates.

Traditional Google ads used fixed templates, limiting customization and failing to capture the unique identity of our clients' brands.

My goal for Superbloom was to create a dynamic system that allowed ads to be tailored more flexibly to the diverse needs of advertisers, whether they were a small local business or a large corporation.

The impact of this project extended beyond Google Display ads and influenced ad creation across platforms, including YouTube, setting a new standard for customizable, high-quality digital advertising.

Context

To understand the scope of this project, it’s essential to first explore how traditional online ads are generated.

How Online Ads Work

As a webpage loads, a space is reserved for Google ads. Through an auction process, an advertiser interested in promoting a product is selected. Google then creates an ad format on behalf of the advertiser and displays it on the website or app.

Ad Templates

To build the ad, Google uses assets provided by the advertiser, arranging them according to a pre-set format template.

Examples of Ads

These ads are highly responsive, adapting seamlessly to various sizes, languages, and devices. They can handle a wide range of asset qualities, text lengths, image aspect ratio, and even cases of missing logo or colors.

Problem Definition

Advertisers were confined to using hard-coded ad templates that offered little flexibility, resulting in ads that looked similar regardless of the business type.

This uniformity didn't effectively communicate the unique value or brand identity of diverse businesses, whether it was a local funeral home or a trendy candy shop.

Different Advertisers

Here are three advertisers with very different brand identity, marketing goals and target audiences.

One Size Fits All Ad Formats

While very efficient and extreamly responsive to their context, traditional ad formats fail to capture the uniqueness of each brand or to tailor the message to each particular audience, resulting in a one size fits no-one type of advertisement.

Superbloom: a new type of ad format

With Superbloom, I aimed to develop a solution that could produce distinct, high-quality ads automatically—respecting the individuality of each business while still leveraging the scalability of Google’s ad infrastructure.
This involved accurately understanding and representing brand identities, enhancing the quality of assets provided by advertisers, and delivering ads that resonated with different audiences, without manual intervention.

Vision and Leadership

My Role

My role was to define and champion the vision for Superbloom, transforming a vague concept into a compelling product vision.

I guided the project through numerous iterations, ensuring that the design and development stayed true to the core objective of creating bespoke ads on the behalf of the advertisers.

1 UX among 15 ENG

I actively participated in strategic planning, setting clear goals, and communicating the vision across teams.
My leadership was instrumental in aligning cross-functional teams, including engineering, product management, and AI research, around a shared understanding of what Superbloom aimed to achieve.

I also provided course corrections when different parts of the team veered off track, keeping everyone focused on delivering a product that balanced innovation with practical advertiser needs.

Vision

Not only engeneering, AI research and product had different individual goals but the overall direction for the project was very obscure since it was the first of its kind, using technology that was developped as we were workign on it. To rally all teams a common goal, I had to create a strong product vision and established with clear yet flexible milestones. Storytelling and visualization were the most efficient tools.

Challenges and Key Decisions

The project faced several significant challenges.

One was how to accurately understand and represent the brand identity of diverse advertisers automatically, without manual intervention.
This involved developing AI models that could interpret brand guidelines, colors, and styles from minimal input.

Another challenge was dealing with low-quality assets provided by advertisers.
Many small businesses lacked the resources to produce high-quality images or videos, which were essential for effective ads.

My team and I decided to focus on enhancing these assets through AI rather than changing the entire ad submission process.
We prioritized developing tools like the uncropping feature, which could intelligently extend image backgrounds, and a tool that enhanced the quality of low-resolution images.

Example of AI feature: smart cropping

Research and Insights

User research played a critical role in shaping Superbloom’s direction.

Through interviews and testing with a diverse range of advertisers, we discovered that small and medium-sized businesses were particularly receptive to AI-driven ad solutions.

These businesses often lacked the resources for professional marketing and were excited about the prospect of automated, high-quality ads.

Conversely, larger corporations, which typically have established brand guidelines and creative teams, were more hesitant to adopt automated solutions, fearing a loss of creative control.

These insights led me to refine the product’s focus, prioritizing features that offered value to smaller businesses while ensuring that larger advertisers could maintain a level of oversight and customization.

Collaboration and Cross-Team Efforts

Effective collaboration across multiple teams was crucial to the success of Superbloom.

I played a key role in facilitating communication and cooperation between teams responsible for different aspects of the project—from front-end design to back-end engineering and AI research.

I participated in sprints with other teams, sharing my expertise in ad formats and ensuring that everyone was aware of the technology we were developing.

I also organized workshops to present the latest tools and features created for Superbloom, fostering a culture of knowledge-sharing.
These workshops not only helped avoid potential roadblocks but also generated interest from other parts of Google, leading to collaborations with teams like YouTube Ads.

Impact and Results

Superbloom introduced several innovative features, such as the uncropping tool, which allowed designers to creatively extend images, and AI-powered enhancements that improved the visual quality of low-resolution assets.

These tools were widely adopted across Google’s ad platforms, enhancing the quality and diversity of ads served to users.

The project also influenced how Google interacts with advertisers, offering them more control over the balance between performance and brand consistency.

For small and medium businesses, Superbloom became a game-changer, enabling them to produce professional-quality ads without needing extensive resources.

Key Takeaways

My experience with Superbloom underscored the importance of vision and storytelling in large-scale projects.

By crafting a detailed, compelling vision from the outset, I ensured that all teams were aligned and focused on a common goal.

Empathy—understanding the perspectives and constraints of different teams—proved invaluable in navigating the complexities of cross-functional collaboration.

The project highlighted the need to balance creativity and performance, particularly in an environment as data-driven as Google.
Keeping the end user in mind throughout the process, despite the technical and business challenges, was key to the project’s success and serves as a model for future work.