data

Snowflake Industry Days: Data Fabric Implementation and Governance at Sumitomo Mitsui Trust Group

Norbert Gehrke

07 Dec 2025 — 10 min read

The Snowflake Industry Days 2025 Japan took place in November, with a Financial Services track session featuring Trust Base and the Sumitomo Mitsui Trust Group, titled "Connecting Sumitomo Mitsui Trust Group: Implementation and Governance of the Data Fabric Concept."

Presenters:

Gen Uehara (Head of Financial Industry, Snowflake)
Satoshi Tanaka (CEO, Trust Base / Senior Researcher, Digital Planning Department, Sumitomo Mitsui Trust Bank)
Daisuke Nakamura (Senior Manager, Trust Base / Researcher, Digital Planning Department, Sumitomo Mitsui Trust Bank)
Takuto Yamamoto (Associate, Trust Base / Digital Planning Department, Sumitomo Mitsui Trust Bank)

1. Executive Summary

This session detailed the strategic initiative undertaken by the Sumitomo Mitsui Trust Group (SMTG) to modernize its data infrastructure through the creation of a "Data Fabric." Led by Trust Base, the group’s digital strategy subsidiary, this project aims to dismantle the rigid data silos inherent in the trust banking sector.

The core challenge addressed is the fragmentation of data across various business lines (banking, real estate, pension management) due to strict legal firewalls and legacy systems. By leveraging Snowflake as the central data platform and ThoughtSpot for business intelligence (BI), Trust Base has established a secure, agile environment where data can be sanitized, shared, and analyzed across the group without compromising security or compliance.

The presentation highlighted the architectural shift from on-premise, waterfall-based development to a cloud-native, agile methodology. Key achievements include the reduction of manual Excel processing, the integration of third-party market data via Snowflake’s marketplace, and the establishment of a governance model that balances "offensive" data utilization with "defensive" data protection.

2. Snowflake Update: The "Easy, Connected, Trusted" Era for Finance

Before the main case study, Gen Uehara from Snowflake provided an industry update, framing the company's evolution from a Data Warehouse to a comprehensive AI Data Cloud under the theme: "Easy, Connected, Trusted" .

2.1 The Complexity of Modern Financial Infrastructure

Uehara opened by acknowledging the immense complexity of the current financial IT landscape. Financial institutions are grappling with data scattered across disparate applications, platforms, and databases. While the goal is automated insight, the reality often involves manual searches across SharePoint files and disjointed databases to find "insights" that are then manually collated . This fragmentation makes it difficult to achieve a holistic view of the business or the customer.

2.2 The Role of AI Agents and Snowflake Intelligence

To address this, Snowflake is positioning itself as the backbone for the "AI Agent" era.

Easy: Snowflake aims to lower the barrier to data interaction. Uehara introduced Snowflake Intelligence, a feature designed to allow users to interact with structured and unstructured data using natural language . This capability is critical for financial advisors who need to quickly query complex client portfolios or transaction histories without writing SQL.
Connected: The value of data multiplies when combined. Uehara emphasized the transition from internal data silos to an interconnected ecosystem. He highlighted the Snowflake User Community in Japan, specifically the financial user group, which has grown to over 70 companies with 160 participants . This community facilitates the sharing of best practices regarding governance and internal approvals, accelerating the industry's collective learning curve.
Trusted: In finance, trust is non-negotiable. AI agents must provide explainable answers with citations, ensuring that recommendations (e.g., revenue projections) are traceable to specific background data . Furthermore, Uehara announced that Snowflake has been registered on the ISMAP (Information system Security Management and Assessment Program) cloud service list, a critical certification for Japanese government and financial institutions to adopt cloud services confidently .

3. Trust Base: The "Moon" Strategy for Banking DX

Satoshi Tanaka, CEO of Trust Base, introduced the unique position of his company within the massive Sumitomo Mitsui Trust Group (SMTG).

3.1 Organizational Context and The "Moon" Analogy

Established in April 2021, Trust Base is a 100% owned subsidiary of the Sumitomo Mitsui Trust Group. Tanaka employed a vivid analogy to explain why a separate entity was necessary for Digital Transformation (DX): The Earth vs. The Moon.

The Earth (The Bank): Represents stability, gravity, and rigid rules. The bank deals with critical assets and must maintain robust, fail-safe operations. It is difficult to "jump high" or move fast in such a heavy-gravity environment .
The Moon (Trust Base): Represents a low-gravity environment designed for exploration and speed. Trust Base was created to be distinct from the bank—physically located in a separate office and operating under different rules—to allow for agile experimentation and rapid prototyping.

The strategic intent is for Trust Base to act as an R&D lab and a hub for specialized talent. They test new technologies and methodologies (like Snowflake and Agile development) in this "low gravity" environment. Once a solution is proven successful, they bring the "knowledge and insights" back to the "Earth" (the main banking group).

3.2 Human Capital Strategy

Trust Base has adopted a hybrid talent strategy. While it includes members seconded from the bank who possess deep domain knowledge, approximately 60% of the staff are external hires with specialized digital skills.

The organization is structured into specialized "Centers" rather than traditional departments:

Business Design Center: Strategy and new business development
UX Design Center: Customer experience and interface design
Data Science Center: Analytics and AI implementation
DX Platform Center: Cloud infrastructure and security
Digital Operation Center: Process optimization and RPA
Software Engineering Center: Application development

This structure allows Trust Base to assemble cross-functional "Squads" for specific projects, ensuring all necessary capabilities are present in a single team.

4. The Challenge: Data Silos in Trust Banking

The core problem necessitating the Data Fabric project is the extreme silo-ization of data within SMTG. Trust banking is a conglomeration of distinct businesses: banking, real estate brokerage, corporate pension management, stock transfer agencies, and asset management.

4.1 Legal and Structural Barriers

Because these businesses handle sensitive client information, they are separated by strict legal firewalls (e.g., banking data cannot easily be shared with the real estate division). Consequently:

Data Isolation: Data is generated and stored within specific business units
Contractual Fragmentation: External data (market data, news feeds) is purchased via individual contracts per department, leading to duplication and inefficiency
Legacy Formats: Data exchange often relies on exchanging massive, password-locked Excel files via email or file servers, making automated analysis impossible
Infrastructure Limitations: Analysis is often performed on local PCs, which lack the computing power to process large datasets, leading to slow insights

Tanaka described the goal as creating a world where these silos are bridged—not by disregarding the law, but by creating a governance layer that allows compliant data sharing.

5. The Solution: Snowflake-Centric Data Fabric Architecture

To solve these challenges, Trust Base architected a Data Fabric. This is not merely a central database but an ecosystem of tools and governance processes that connects data providers with data consumers.

5.1 Technical Architecture

The architecture is built primarily on Snowflake (on AWS) for data storage and processing, and ThoughtSpot for the visualization and analytics layer.

The data flow is designed to ensure security while maximizing accessibility:

Ingestion: Data flows from SMTB (the bank) and other group companies into the Data Fabric. This includes both structured database records and unstructured Excel/CSV files.
Sanitization and Access Control: This is the most critical step. Data does not flow freely to everyone.
- Secure Views: Data is wrapped in "Secure Views" within Snowflake to control visibility at a row and column level.
- Project-Specific Accounts: To prevent cross-contamination, Trust Base creates dedicated Snowflake accounts for specific projects (e.g., "Case A Account," "Case B Account"). Data is shared into these accounts using Snowflake's Data Sharing capabilities rather than copying the data (ETL), which reduces latency and storage costs.
Visualization: ThoughtSpot is connected to these specific Snowflake views, allowing users to query data using search-based analytics.

5.2 Why ThoughtSpot?

Nakamura explained a pivotal decision regarding the BI tool. Initially, the group planned to use a different, major BI tool widely used in the banking sector. However, during the PoC (Proof of Concept) phase in April 2025, they encountered a significant hurdle: Licensing Models .

The incumbent BI tool used a user-based licensing model. To achieve the goal of "democratizing data" across the entire group (potentially tens of thousands of employees), the cost would have been prohibitive. Furthermore, the incumbent tool required specialized "analyst" skills to build dashboards, which created a bottleneck.

Trust Base pivoted to ThoughtSpot because:

Licensing: It offered a capacity-based model (based on data usage/queries rather than seat count), making it scalable for broad organizational rollout.
Usability: Its AI and search-driven interface meant that non-technical users in business departments could answer their own questions without waiting for the data team to build a dashboard.

5.3 "Offense" and "Defense" in Governance

The architecture explicitly balances two opposing needs:

Defense (Data Protection): Preventing leakage, ensuring compliance with banking laws, and managing access rights. This is handled by the "Data Governance Office" and Snowflake's rigorous role-based access controls (RBAC).
Offense (Data Utilization): Enabling marketing analysis, new business creation, and operational efficiency. The platform allows users to safely access sanitized data to find cross-selling opportunities or market trends.

6. Agile Implementation Methodology

A significant portion of the session detailed how the team built this platform. Breaking away from the bank's traditional "Waterfall" development style, Trust Base adopted Agile (Scrum).

6.1 Why Agile?

Yamamoto explained that in data projects, requirements are rarely fully understood at the start.

Waterfall Risk: In a traditional model, specifications are fixed upfront. If the business realizes halfway through that they need different data, it is often too late or expensive to go back.
Agile Advantage: By using 2-week sprints, the team could release a small version of the dashboard, get immediate feedback from the business users ("This graph is wrong," "I need this column"), and iterate.

6.2 Workflow Tools

The team uses Asana to manage their backlog and tasks visually. The cycle involves:

Sprint Planning: Deciding what to build in the next 2 weeks.
Development: Configuring Snowflake pipes, cleaning data, building ThoughtSpot worksheets.
Sprint Review: Showing the "Increment" (working software) to the user for feedback.
Retrospective: The team discusses what went well and what to improve.

This approach allowed them to pivot quickly. For example, the switch from the initial BI tool to ThoughtSpot was a major architectural change that would have been catastrophic in a Waterfall project but was manageable in Agile .

7. Use Cases

The session highlights three specific use cases that demonstrate the versatility of the Data Fabric.

7.1 Use Case 1: Trust Base Service Log Analysis (Internal)

Status: In Production.
The Problem: Trust Base operates various digital services. They needed to analyze user behavior logs to improve UX.
The Solution:
- Logs are exported from the service's database (DB) to Amazon S3 (compressed via gzip).
- Snowflake ingests these logs into an "Internal Coordination Account."
- Data is shared to a "Shared Account" via Secure Views.
- ThoughtSpot connects to this Shared Account.
Benefit: Developers and Product Owners can analyze user behavior in real-time without needing to write SQL or build complex ETL pipelines. This supports a DevSecOps culture where feedback loops are fast.

7.2 Use Case 2: Escaping the "Excel Hell" (3rd Party Data)

Status: PoC (Proof of Concept).
The Problem: A contracting department at SMTB receives market data from a third-party provider in the form of 3,000 Excel sheets.
- Formats vary wildly.
- Files are heavy and crash local PCs.
- Analyzing trends across multiple sheets is impossible.
The Solution:
- Step 1 (Manual): Trust Base manually ingests these Excel files into Snowflake tables.
- Step 2 (Future Automation): They are renegotiating the contract to have the vendor provide data directly via Snowflake Marketplace or automated data sharing.
Benefit: What used to be disparate files are now queryable tables. Users can search "Bond Yield Trends 2024" in ThoughtSpot and get an answer derived from thousands of underlying sheets instantly.

7.3 Use Case 3: Integration of 3rd Party and User Data

Status: In Progress.
The Problem: The group buys data from providers (like Bloomberg, MSCI, etc.). Currently, Department A buys it and keeps it locally. Department B might need it but can't access it.
The Solution:
- Data Providers share data directly to a Trust Base "External Coordination Account" on Snowflake.
- Trust Base performs "Cleansing" and "Name Integration" (matching external company names with internal client IDs).
- This enriched data is shared with specific business units based on their contract rights.
Benefit: This creates a "Golden Record" where external market intelligence is mapped directly to internal customer portfolios, enabling highly targeted marketing and risk analysis.

8. Future Roadmap and Challenges

The project began construction in April 2024 and launched its first phase (non-PII data) in October 2025. The roadmap extends into 2026 with critical developments.

8.1 Handling PII (Personal Identifiable Information)

Currently, the platform mostly handles anonymized or general business data. The next major milestone (Phase 2) is to create a secure environment for Personal Information.

This requires strict "Sanitization" logic.
They are designing a system where the "Data Owner" (the bank department) can define masking levels (e.g., "Mask Name," "Hide Address") before the data is shared with other departments.

8.2 Establishing the Data Governance Office

Technology is not enough. Trust Base is establishing a formalized Data Governance Office. This body will:

Define data quality standards.
Manage the "Data Catalog" (so people know what data exists).
Enforce security policies regarding who can see what.

8.3 Expanding the "Service Domain"

Trust Base aims to evolve from a cost center (building internal tools) to a value generator.

Synthetic Data: They are exploring the creation and sale of synthetic data (artificial data that mimics real statistical properties without compromising privacy) as a new business line.
Group-wide Service: Expanding access beyond the bank to other group companies like Sumitomo Mitsui Trust Asset Management.

9. Conclusion

The "Data Fabric" initiative at Sumitomo Mitsui Trust Group represents a significant departure from traditional banking IT projects. By utilizing Trust Base as a "Moon" base for innovation, the group has successfully deployed a modern, cloud-native stack (Snowflake + ThoughtSpot) using Agile methodologies that are typically difficult to execute in a regulated banking environment.

The shift from file-based data exchange to a Snowflake-based Data Sharing architecture addresses the fundamental inefficiencies of the trust banking sector. It transforms data from a static, siloed asset into a fluid, queryable resource. As Yamamoto concluded, the ultimate goal is to create an environment where data is not just "guarded" for compliance, but actively "trusted" and shared to drive new business value, turning the Trust Group into a data-driven leader in the financial industry.