Applying design system principles to make Data Mesh usable, scalable, and human-centered.

Data Signals


## Overview
As Grab continued expanding across Southeast Asia, our data ecosystem grew exponentially. Every new business line; ride-hailing, deliveries, financial services, added new pipelines and metrics. 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 our data to be reliable, contextual, and actionable, not just abundant.

This transformation demanded more than infrastructure. It required design leadership to make governance usable, visible, and scalable.

This is the story of how we built Signals Marketplace, Grab's internal Data Mesh, and how design shaped the systems, principles, and collaboration models that turned data trust into an organizational capability.

### What is Data Mesh?
In the broader data world, Data Mesh is a decentralized approach to data architecture. Instead of a single, centralized data warehouse, ownership is distributed across domain teams, each treating data as a product with its own lifecycle, quality standards, and documentation. The model emphasizes four core principles: domain-oriented ownership, data as a product, self-serve infrastructure, and federated governance. Together, these principles make large-scale data ecosystems more scalable, reliable, and adaptive, especially in organizations with diverse business lines like Grab.

## Designing for Scale and Trust

### Role
I lead Product Design, working with the Head of Product and Heads of Engineering to define Grab's data transformation vision. As the initiative grew, I led a multidisciplinary design team across the data platform, including catalogs, monitoring, and discovery. The work sat at the intersection of data infrastructure and product experience, ensuring teams could understand and trust data within their workflows.

My role evolved through three phases:

1. **Foundation:** Defined the design strategy and principles for how trust, certification, and ownership surfaced across tools.  
2. **Coordination:** Aligned product, engineering, my design around consistent principles.  
3. **Enablement:** Coached designers to own domains and contribute to shared governance, ensuring the system evolved beyond any one team.

This work treated data trust as an organizational design challenge.

### Background
Grab's rapid expansion created a sprawling, interconnected data ecosystem. With each new business vertical came new pipelines and metrics, making it increasingly difficult to maintain alignment and consistency. As the company began investing heavily in AI and automation, it became clear that scale alone wasn't enough. Data had to be dependable, discoverable, and understood. Without that foundation, every model, forecast, and decision risked being built on uncertainty, undermining the very intelligence we aimed to unlock.

### Challenges
In 2024, we launched Signals Marketplace, Grab's Data Mesh initiative, to decentralize data ownership and rebuild trust at scale. Three challenges shaped our approach:

1. **Scale and fragmentation:** Each business line produced its own pipelines and definitions, creating overlap and inconsistency. Teams duplicated work to maintain control.  
2. **Centralization and unclear ownership:** A central engineering model had become a bottleneck. Producers and consumers operated in silos, with limited accountability for data quality or documentation.  
3. **Uneven literacy and visibility:** Teams varied in data maturity. Some had engineers; others didn't. Many lacked visibility into reliability or lineage, forcing each to rebuild trust from scratch.

We needed a model that scaled trust as effectively as access, and that realization led us to reimagine data ownership through Data Mesh.

### Solution
We introduced Data Mesh, treating data as a product owned by domain teams. This empowered those closest to the data to guarantee its quality, reliability, and accessibility.

Our goal: make trust visible and measurable. We designed systems that let teams define, verify, and communicate data quality directly.

<DataTrustFlow />

### Certification: The Cornerstone
Certification made trust tangible. It was both a technical mechanism and a design component, applied consistently across data surfaces.

- **Purpose:** For producers, certification formalized guarantees such as schema, freshness, and quality standards. For consumers, it provided a visible trust signal, reducing manual validation.  
- **Data contracts:** Each certification included a data contract defining SLAs, schema rules, and change protocols, establishing accountability between producers and consumers.  
- **Ownership clarity:** Every certified asset had Business Data Owners (BDOs) and Technical Data Owners (TDOs) accountable for maintenance and reliability.

For the first time, teams could see which data they could rely on, moving from assumption to assurance. Together, certification and contracts created a framework for designed trust, where clarity replaced ambiguity and reliability scaled with confidence.

<CertificationCard />



## Treating the Problem as a Design System

We structured the system around three principles: what we built, how we scaled it, and how we made it actionable.

To operationalize Data Mesh, we treated it as a design system built from modular components, sustained through contribution, and designed for action.

Each designer owned a data domain (events, attributes, metrics, or datasets) and was accountable for applying certification and contract components within their context. The certification indicator conveyed trust signals such as ownership, data quality, and reliability, applied consistently across experiences. Designers ensured the components met their domain's needs and, when gaps appeared, contributed updates to align on broader adoption.

**Components:** At the core were reusable, governed components ensuring consistency and clarity across data products. The primary element was Certification, a pattern embedded across platforms and assets. Designers refined its visuals and interaction model to create a unified language of trust consistent across data products. Certification defined how ownership, freshness, and quality signals appeared, ensuring every data product communicated reliability in a clear, standardized way.

Complementing it were Data Contracts, which defined the rules and checks behind certification. They powered trust indicators and kept certification a live reflection of data health.

<DataProductTabs />

**Contribution:** To accommodate diverse data assets such as metrics, events, attributes, and datasets, we built a contribution model balancing flexibility with governance. Designers across domains extended or refined patterns to fit their context while maintaining coherence through shared reviews and documentation. This ensured every contribution strengthened the system rather than fragmenting it. It also fostered ownership and collaboration, uniting design and data teams around a shared visual and behavioral language.

**Actionability:** The system was designed for action, not just visibility. When data contracts detected anomalies or breaches, the platform surfaced contextual alerts and clear ownership paths. These proactive signals guided teams toward resolution instead of reaction.

## Future-Proof Foundations

Within months, the impact was both technical and cultural:

- **75% of queries targeted certified assets**, demonstrating trust and adoption.  
- **Redundant datasets dropped 50%**, improving discoverability and maintenance.  
- **A 400% year-over-year rise in the deprecation of low-quality tables** showed teams cleaning up proactively.  
- **Ownership visibility** improved, linking every certified asset to accountable teams.

This effort reshaped how Grab worked with data, reducing noise, surfacing signals, and embedding trust into daily workflows. It changed how teams thought about data.

### Key Learnings
Adoption required technical clarity and behavioral change. We showed teams why certification mattered. It added effort but delivered cleaner pipelines, faster debugging, and shared confidence that accelerated decision-making.

As AI takes a larger role in decision-making, context defines value. Reliable, contextual data ensures every insight and model is grounded in truth. This initiative built the foundation for that future.

### Reflection
My role evolved from contributor to orchestrator. I guided designers to move from interface design to system enablement. I helped designers gain confidence in navigating technical complexity, ensuring they saw themselves as equal partners in shaping data systems, thinking in terms of relationships, governance, and adoption. Our role was to design conditions for trust and collaboration.

At scale, design's role is to shape how organizations see, not just how they look. Leading design in technical systems means uniting logic with empathy, and structure with adaptability. The Signals Marketplace proved how design systems thinking can scale trust across an organization. The real success wasn't just cleaner data, it was teams learning to trust what they build together.


Loading Project...