Making complex audience building clearer, reusable, and easier to manage.

The audience management solution in Leanplum that helps marketing teams build, manage, and better understand their diverse audience and leverage this data in their marketing strategy.

Audience Management

Leanplum is a leading multi-channel customer engagement platform that helps forward-looking brands meet their customers' real-time needs. Marketers think in the context of Audiences, yet Leanplum was providing a User-level view, which was too granular for the marketers' needs.

Overview

Leanplum emphasized individual users over audience segments. Saving an audience was technically possible, but hidden in a hard-to-discover dropdown, leading to repeated manual work.

Product Analysis

Only insight available was audience size - no channel preferences, health indicators, or behavior breakdowns. Key actions were hidden, leading to frustration and errors.

User Pain Points

Most marketers didn't know you could save an audience. Even those who did rarely used the feature - they didn't trust it due to lack of guardrails and visibility.

User Interviews

Competitors had audience management tools, but their solutions lacked depth, especially around reuse, collaboration, and insights. We saw this as an opportunity to differentiate.

Competitive Analysis

Forrester and Gartner validated our hypothesis that marketers think in the context of Audiences, and they want to be empowered in the way they can define them.

Analyst Reports

We conducted comprehensive research to understand how marketers managed audiences and identified critical gaps.

Research

We needed to shift from a user-level mindset to one centered around Audiences as first-class citizens. The new experience had to elevate Audiences as a primary concept, allowing marketers to view audiences holistically, validate their structure, and preview representative user samples.

Insights

Make Audiences a first-class citizen in the interface, creating a dedicated, centralized space that laid groundwork for more scalable strategic workflows.

Centralization

Prioritize discoverability and collaboration, making it easy for teams to access, edit, and engage audiences while reducing repeated work.

Productivity

Create a consistent experience for building, saving, and engaging with audiences, ensuring users could apply their knowledge across different parts of the product.

Consistency

Embed actionable insights at every level, helping marketers understand their audience and choose the right engagement strategies with health metrics and engagement data.

Meaningful Insights

Design the UI to support exploration and iteration. Users could experiment with filters, preview results, and revert changes, lowering the risk of mistakes.

Flexibility & Forgiveness

Five core principles guided our design approach to transform audience management.

Design Approach

Returned to users with early concepts to validate direction, uncover edge cases, and ensure the updated experience wouldn't disrupt existing workflows.

Regular critiques and team reviews kept us aligned across design, product, and engineering, surfacing implementation challenges early and sparking ideas for deeper functionality.

Design Critiques

Phased beta with engaged users revealed friction points. We thoughtfully reintroduced individual user search one layer deeper to provide continuity while guiding toward strategic workflows.

Beta Program

We validated our approach through continuous iteration and user feedback at every stage.

Feedback and Iterations

We launched a phased beta with engaged users. A key friction point was finding individual users, a familiar habit from the old product. This was a deliberate shift to an audience-first model to scale campaigns and reduce repetition. To ease the change, we brought back user search within the Audiences view, preserving continuity while guiding users toward a more strategic workflow. Some found it disorienting at first but later saw the value in a broader perspective.

Beta Program Insights

A dedicated dashboard for all saved audiences with organized, searchable views and real-time metrics. Teams get organized views with metadata and real-time metrics showing where each audience is being used, plus high-level health insights across the user base.

Audience Dashboard

The core space for building and refining audiences with powerful yet intuitive tools. Designed around Clarity, Flexibility, and Forgiveness with drag-and-drop rule rearrangement, merge/split logic blocks, and undo/redo for confident exploration.

Audience Builder

Detailed user sampling view for transparency and confidence in audience logic. Preview real users in any audience, validate audience logic matches expectations, and provide transparency into actual user data for quality assurance.

Users in Audience

Audience insights enabled smarter segmentation and more relevant messaging. By helping marketers understand audience breakdown and behavior patterns, we empowered more informed campaign decisions.

Campaign Impact

Reusable audiences significantly reduced campaign setup time. With visibility into existing audiences, marketing teams avoided redundant segmentation work and aligned more quickly.

Operational Efficiency

The solution encouraged collaboration between teams, ensuring visibility and promoting knowledge sharing. It served as a guardrail preventing duplication or over-communication.

Team Collaboration

The new audience management platform delivered measurable improvements across multiple dimensions.

Impact

Audiences laid the foundation for future innovation in targeting, experimentation, and automation. This was the first step in reimagining how marketing teams understand and engage with their users. Many solutions feel obvious in hindsight, but took rounds of iteration and collaboration to achieve.

Reflection

Making trusted data easier to discover, understand, and use across a complex platform ecosystem.

Applying design system principles to make Data Mesh usable, scalable, and human-centered. Building Signals Marketplace, Grab's internal Data Mesh, to turn data trust into an organizational capability.

Data Signals

Defined the design strategy and principles for how trust, certification, and ownership surfaced across tools.

Foundation

Aligned product, engineering, my design around consistent principles.

Coordination

Coached designers to own domains and contribute to shared governance, ensuring the system evolved beyond any one team.

Enablement

Role

As Grab expanded across Southeast Asia, our data ecosystem grew exponentially. Access to data was improving, but the ability to trust and use it confidently was not. To move toward an AI-driven future, we needed data to be reliable, contextual, and actionable.

The Challenge

Each business line produced its own pipelines and definitions, creating overlap and inconsistency. Teams duplicated work to maintain control.

Scale and Fragmentation

A central engineering model had become a bottleneck. Producers and consumers operated in silos, with limited accountability for data quality.

Centralization Bottleneck

Teams varied in data maturity. Many lacked visibility into reliability or lineage, forcing each to rebuild trust from scratch.

Uneven Literacy

Three challenges shaped our approach to rebuilding trust at scale.

Three Core Challenges

We introduced Data Mesh, treating data as a product owned by domain teams. Our goal: make trust visible and measurable through systems that let teams define, verify, and communicate data quality directly.

Making Trust Visible

Certification made trust tangible. It was both a technical mechanism and a design component, applied consistently across data surfaces. For producers, it formalized guarantees. For consumers, it provided visible trust signals, reducing manual validation.

Certification: The Cornerstone

We treated Data Mesh as a design system built from modular components, sustained through contribution, and designed for action. Each designer owned a data domain and was accountable for applying certification and contract components within their context.

Design System Approach

Demonstrating trust and adoption across teams.

75% of queries targeted certified assets

Improving discoverability and maintenance.

50% reduction in redundant datasets

Teams proactively cleaning up low-quality tables.

400% year-over-year rise in deprecation

Every certified asset linked to accountable teams.

Ownership visibility

Within months, the impact was both technical and cultural.

Adoption required clear technical standards and behavior change. Certification took more effort, but it led to cleaner pipelines, faster debugging, and better decisions. As AI expands, trustworthy data becomes critical.

My role shifted from doing the work to enabling others. I helped designers move beyond interfaces to shaping systems, giving them confidence to work with technical complexity and data governance.

The Signals Marketplace showed that design can scale trust. The real win was not just better data, but teams trusting what they build together.


Key Learnings & Reflection

Giving teams greater visibility and control over the data powering their products and campaigns.

The solution that helps data governors manage the process of sourcing, analyzing, and controlling the available data in Leanplum to leverage in their business and marketing strategies.

Data Management

Making Data Accessible and Manageable for Marketing Teams. While Leanplum's messaging and engagement tools were powerful, data capabilities had a critical gap. Data was technically accessible, but discovering, managing, or activating it often required developer support, creating bottlenecks for marketing teams.

The only data-facing tool in the platform was a static Events list with no metadata, context, or controls.

Static Events List

User attributes lacked a dedicated management interface, and importing/exporting data was primarily done through APIs.

No Attribute Management

CSV uploads required developer intervention. Partner integrations existed, but weren't visible or configurable from the UI.

Developer Dependencies

Non-technical users, especially marketers and CRM leads, were regularly blocked from executing basic workflows.

User Workflow Blocks

Our support team spent a disproportionate amount of time responding to tickets related to data imports, partner setups, and event discovery.

Support Volume

The only data-facing tool in the platform was a static Events list with no metadata, context, or controls. During early audits and user interviews, it became clear that users lacked a clear source of truth for their data.

Problem

We audited the existing experience and reviewed common support tickets. The Events list presented raw data, but lacked the metadata and control users needed to make sense of it.

Marketers often shared that they weren't confident in what data existed or how reliable it was. Many avoided attempting data uploads or partner integrations altogether to avoid back-and-forth with engineering.

Support Ticket Review

Setting up CSV imports or integrations required combing through documentation or working with developers. In practice, many marketers simply avoided these workflows altogether.

User Workflow Analysis

We began by auditing the existing experience and understanding user needs through comprehensive analysis.

Data Management - Central Hub for All Data Operations. We introduced a new top-level navigation item, Data Management. This became the central hub for all data integrated into and out of Leanplum. I worked closely with the PM and engineering leads to define the information architecture and interaction model for this new surface.

Solution

View and edit all tracked user events, with added metadata and validation controls.

Events

Manage user properties through a similar interface to Events.

Attributes

A directory of third-party integrations with setup flows embedded directly into the UI.

Partners

A drag-and-drop interface for non-technical users, with inline validation and upload history.

CSV Uploads

We built four modular, interconnected components. Each module reused consistent patterns from the rest of the product.

Four Core Modules

Initial Design Approach. We initially explored using a two-level detail view pattern similar to what we had implemented in the Audiences page. However, due to technical constraints, saving edits on individual records would block the entire interface while the backend job completed. This would have introduced a frustrating delay for users.

Events & Attributes Design

User Testing Reveals Critical Issues. We tested early flows with a small beta cohort and used their feedback to refine how we surfaced data status. One of the most significant insights came when users struggled with the Events module's save behavior. Many were confused when editing individual events caused delays or interface blocking. This validated the need for our batch-edit single-page model and clarified our prioritization going forward. The feedback reinforced that, while the adjustment required a learning curve, it ultimately aligned better with how teams work across multiple events in one session.

Beta Validation

From Individual Edits to Batch Workflows. Based on beta feedback, we optimized batch workflows to meet user expectations at scale and shifted additional design focus toward clarity, responsiveness, and feedback mechanisms for these batch actions.

Refined Solution

Final Single-Page Layout. To resolve the interface blocking issue, I designed a single-page layout that allowed users to make batch edits across events and then save them together. This pattern preserved flexibility while accommodating technical constraints. We added persistent save indicators and inline validation to prevent surprises.

Events Management

Dedicated Interface for User Properties. User attributes gained a management interface similar to Events. Marketing teams could now manage customer properties directly without relying on documentation or developer support.

User Attributes Management

Making Integrations Discoverable and Actionable. Integrations were previously handled through engineering teams and documentation. We made them discoverable and actionable in-product by introducing a Partners page where users could view and configure integrations directly from the UI. This also helped surface and promote the full set of available integrations in Leanplum, many of which users were previously unaware of.

Streamlined Drag-and-Drop Experience. We designed a streamlined drag-and-drop experience for uploading user data via CSV. Integrated validations provide immediate feedback on file structure and formatting issues before upload. To support collaboration and traceability, we also included a visible upload history with metadata such as filename, upload time, and uploader.

Non-technical users gained the ability to access and manage data without relying on engineering or support. This unlocked faster iteration cycles and gave marketing teams greater autonomy in campaign setup.

User Empowerment

Tickets related to data access, configuration, and imports dropped significantly, allowing support teams to focus on higher-impact work.

Support Load Reduction

By surfacing and enabling partner integrations directly within the platform, we extended Leanplum's capabilities and positioned Data Management as the foundation for future ecosystem expansion and vendor flexibility.

Expanded Product Offerings

Data Management became the foundation for broader data governance and user behavior segmentation. By centralizing data visibility and control, we enabled more sophisticated campaign strategies.

From Blocker to Strategic Enabler. Data Management transformed how Leanplum treated data-no longer just a backend artifact, but a strategic, user-facing capability that became a core enabler of campaign strategy and lifecycle marketing.

A lightweight design system with essential UI components and responsive layouts.

Enabling product teams to plan, launch, monitor, and learn from experiments with greater confidence.

The experimentation platform supports how teams discover, run, evaluate, and roll out experiments at scale. As usage grew, the platform needed to evolve into a system that accelerates learning, improves decision-making, and reduces the cost of iteration.

A redesigned dashboard optimized for speed, clarity, and scale.

Interactive Experiment Dashboard

Experimentation Platform

I led Product Design for the experimentation platform, owning strategy, discovery, and execution end to end.

Understood how experimentation worked end to end across discovery, evaluation, and setup by analyzing usage data and collaborating closely with engineers, data scientists, and product managers.

Product Discovery

Translated insights into a clear product strategy and roadmap, prioritizing improvements across the Dashboard, Experiment Details, and Creation Flow.

Product Strategy

Led design from prototyping through implementation, partnering closely with engineering to ship changes that materially improved experimentation velocity and reliability.

Design & Delivery

Experimentation is a core driver of product innovation, but at scale, the tools supporting it began to slow teams down.

Friction accumulated across discovery, evaluation, and setup. The problem was not isolated usability issues. The experimentation workflow itself did not scale.


The platform struggled to keep up as experimentation volume increased, slowing down routine workflows.

Performance and Efficiency

Critical signals were difficult to scan and reason about, increasing cognitive overhead during discovery and evaluation.

Clarity and Information Access

Key workflows introduced unnecessary cognitive load and surfaced errors too late, reducing trust in experiment quality.

Workflow Complexity and Reliability

Three systemic challenges shaped how we approached rebuilding the experimentation platform.

The entry point to experimentation, redesigned for speed and control. A lightweight, scan-friendly dashboard that supports fast discovery, clear ownership, and efficient monitoring at scale.

Optimized for fast discovery and everyday control as experimentation scales.

Experimentation Dashboard

The decision surface for experimentation. A consolidated results view that brings context, metrics, and outcomes into one place, enabling faster evaluation and more confident rollout decisions.

Clear hierarchy and standardized patterns for reliable decision-making.

Experiment Details Page

Experiment Details

The foundation of experiment quality. A guided creation flow with real-time validation and reusable configurations that improves setup reliability and reduces errors.

Structured setup with validation and cloning to support safe, repeatable experimentation.

Treatment Configuration Table

Experiment Creation Flow

Teams could move from idea to outcome more quickly.

Faster Learning Cycles

Teams ran more valid experiments in parallel with the same resources.

Higher Experimentation Throughput

Clearer results enabled faster rollouts and earlier termination of underperforming experiments.

Greater Decision Confidence

By reducing friction across discovery, evaluation, and setup, the platform shifted from slowing teams down to enabling faster, more reliable experimentation.

Experimentation tooling must be treated as a system. Optimizing individual features without addressing end-to-end flow creates local improvements but global friction.

Internal platforms shape how teams work, decide, and allocate effort. Thoughtful design increases velocity and confidence, while poor design quietly taxes the organization.

Product success depends on adoption. Clear communication, onboarding, and change management were essential to ensuring these improvements translated into sustained behavior change.


A kanban-style workflow that turns your Gmail and Outlook inbox into a todo app using inbox zero principles, helping you manage your email and tasks in one central place