BMW Group and Amazon Cloud Technology announced a comprehensive strategic cooperation in 2020. The goal of this collaboration is to further accelerate the pace of innovation at the BMW Group by placing data analytics at the center of decision-making. A key element of this cooperation is the further development of the BMW Group's Cloud Data Hub (CDH). This is the central platform for managing company-wide data and data solutions in the cloud. At re:Invent 2019, BMW and Amazon Cloud Technology showcased a new cloud data center platform. First, they briefly introduced different data platform prototypes, and then introduced the process of building the BMW Group cloud data center.
Solution overview
In the context of regulatory reporting, BMW Financial Services processes critical financial services data containing personally identifiable information (PII). It needs to provide an in-depth analysis of our financial data to one of the European national regulators on a monthly basis, and also needs to comply with Schrems II and GDPR regulations when processing PII data. This requires PII to be pseudonymized when it is loaded into the cloud data center and must be further processed in a pseudonymized form.
To meet these requirements precisely and efficiently, BMW Financial Services decided to cooperate with Amazon Data Lab. Amazon Data Lab has two services: Design Lab and Build Lab.
Design Lab
Design Lab is a 1 to 2 day event for clients who need real architectural advice but are not yet ready to build, based on Amazon expertise. Using BMW Financial Services as an example, before starting the build phase it was critical to bring all stakeholders together to document the various parties (from the owners of the various data sources to the end users who will use the platform for analysis and gain business insights etc.) all functional and non-functional requirements that may affect the data platform. Within the scope of the Design Lab's work, there are three use cases:
Regulatory Reporting: The most important task for BMW Financial Services is the regulatory reporting use case, which involves the collection and calculation of data and reports for filing with national regulators.
On-premises data warehouse: For this use case, all key performance indicators (KPIs) and key value indicators (KVIs) that will be defined during the project need to be calculated and stored. Historical data needs to be stored, but the pseudonymization process needs to be applied in compliance with GDPR directives. Additionally, historical data must be accessed through Tableau visualization tools on a daily basis. Regarding the structure, it is necessary to define two levels (at least): one is the contract level, which is used to justify the calculation of all KPIs, and the other is the aggregation level, which is used to optimize the fix. Use of personal data is restricted within the application, but re-identification must be possible for authorized usage patterns.
Accounting Details: This use case is based on BMW's accounting tool IFT, which provides accounting balances from all local market applications at the contract level. It must be done at least once a month. However, if during settlement some issues are found on the IFT, it must be possible to restart and delete the previous results. After month-end closing is complete, this use case must retain the last version of the accounting balance generated for the current month and store it. At the same time, all versions of accounting balances must be accessible to other applications for inquiries and be able to retrieve information up to 24 months old.
Getting Ready for Build Labs
Typical preparations for a Build Lab following a Design Lab include identifying some examples of common use case patterns, often more complex ones. To maximize Build Lab success, reduce the long-term target architecture to a subset of components that meet the requirements of these examples and can be achieved in intense 3- to 5-day sprints.
For Build Lab to be successful, any external dependencies, such as network connections to data sources and targets, also need to be identified and resolved. If this is not feasible, meaningful ways to simulate these situations need to be found.
Build Lab
The BMW team set 4 days for their Build Lab. During this time, their dedicated Data Lab Architect joined hands with the team and helped them build the following prototype architecture.
