ING, a global bank headquartered in the Netherlands provides retail and wholesale banking services to customers in over 40 countries. The bank’s data analytics team plays a crucial role in informing decision-making, improving customer experience and optimizing business performance.
However, ING’s analytics data architecture employed a traditional monolithic approach with centralized ownership. To efficiently cope with the increasing number of data use cases, ING needs to modernize its data services. A Data mesh implementation would allow ING to decentralize the data ownership from a central team to relevant business domains, which will help ING to scale faster and drastically improve time to market.
Thoughtworks partnered with ING to implement a data mesh proof of concept (PoC) over Google Cloud Platform, including a consumer aligned data product. The PoC was to demonstrate how data mesh could modernize data management and speed up the time-to-market for data products. The PoC was on a real use case and actual production data schema but using synthetic data. This PoC would then serve as a reference for the data transformation journey at ING.
The use case
Over the 8-week project, we collaborated with ING’s Data Management Tribe (DMT), the bank’s data experts.
Together, we selected the use case of customers who started their journey with chatbot and later moved to other channels like human chat or call to resolve their grievance. To reduce operational costs, ING’s goal is to fully resolve customer grievances so that they do not have to move to other channels. The real use case connected business teams and demonstrated how a use case implementation would look with data mesh.
How we created the data product blueprint
The first step of our data mesh approach was to create a data product blueprint targeting their future cloud platform - GCP. It captured all essential information about the data, such as its origin, format, and application, to help users understand and manage it effectively. This allowed the organization to standardize their data products and streamline related services and resources. The blueprint also opens up the possibility to automate data governance to ensure data quality, security and compliance, increasing consistency and reducing manual effort.
To help users comprehend and replicate each process as required, we made the data processing patterns visible. Facilitating discoverability enabled users to locate and access required data with ease.
We used a cross-functional Thoughtworks team including a solution architect, a data engineer and a business analyst. These different perspectives and skills efficiently delivered a functional Proof of Concept, with a well structured narrative, that ING can use to demonstrate the benefits of Data Mesh applied in a concrete realistic case.
What data mesh achieves
Since data mesh promotes decentralized data ownership, teams have the autonomy to develop and manage their data products. They can tailor applications or services to specific use cases without constant approval from a central authority.
Data mesh can handle large quantities of data so teams can scale their projects according to each use case requirement. Since data mesh is self-serve, teams can efficiently respond to changing needs, experiment with new ideas and implement solutions quickly.
The demo by Thoughtworks helped us to clearly understand how creation of data products by autonomous business domain teams & processes around it would look like in a data mesh world with drastically improved time-to-market for data products.
The outcome
The PoC demonstrated the benefits of a data mesh approach, validated key concepts and gave ING with a comprehensive understanding of data management under this model. It provided a reference for how domain teams can autonomously build data products with a self-serve data platform and federated computation governance. This approach will address some of the current challenges faced by ING, including scalability and centralized data, while accelerating time to market for data products.
Although the first PoC was developed over Google Cloud, since the financial industry is highly regulated it requires some data products to be deployed on-premise. The next step involves creating an on-premise PoC for ING that combines having on-premise and cloud based data products visible through a single data catalog, with unified lineage and metadata.
Having real data products as a PoC helps to explain data mesh as a concept, but in order to scale-up their implementation in production other aspects including federated data governance and organizational transformation (including change management) will be part of future work streams with ING.
