โ11-19-2021 12:00 AM
DCU has grown significantly over the past few years, but their systems and integration had not evolved at the pace needed to support this growth. Because of this, their 40+ applications could not present the โSingle Version of the Truthโ, restricted the ability to develop a cost-effective Self-Service Reporting capability and restricted needed Data Governance capabilities.
The legacy environment also had a lack of automation needed to support the scale desired. The architecture was not capable of supporting the data and process scale required to sustain the goals. Lastly, there was extensive manual data manipulation being performed by multiple departments often leading to disparate results.
The customer use cases were:
โข Predictive Analytics - liquidity
โข Near real-time reporting focused on fraud
โข Proactive collections and member stratification,
โข Trend analysis leveraging accurate models
The challenges including un-usable data due to the silos and disconnected systems. THe customer struggled to integrate data or compare across source systems. They also experienced slow, inconsistent, underwriting decisions hampered by data constraints.
Digital Federal Credit Union
EXL Service was engaged to support the goals of DCU: to continue to grow membership as well as grow member products, services & offerings to improve member value. To achieve these goals, we began a multiple phase business and technical analysis of the current environment & operations at DCU. These efforts focused on the following workstreams:
โข Deposits and bank statements
โข Credit Card transactions
โข Early risk predictions
โข Attrition models
The project was started with two initial use cases
โข Executive Dashboard - focused on monitoring key metrics regarding member activity, deposit analysis, product growth, and key success factors
โข Transaction Data Mart - integrated account transaction history providing member behaviour analysis, with identification of product and service opportunities
The management roadmap goals included:
โข Improve ability to deliver additional Self-Service Capability
โข Focus on Data Quality, Data Definition and Data Ownership
โข Need for more integrated data to provide better insights (based on historical data)
โข Goal is to enable a Data Driven Culture
From these findings, we proposed a significant modernization effort to address the issues identified. This effort included the design and development of a Cloud-based Modern Data Platform. As DCU was an existing AWS customer, we developed the solution on AWS using both Amazon and partner technologies. This platform will address the current issues, plus provide the sustainable, scalable technology base needed for the next phase of growth.
Together with SnapLogic we proposed an archtiecture that leveraged SnapLogic with Amazon Redshift for the core capabilities. SnapLogic was able to demonstrate their maturity, performance and features that drove the automation, scale and managed processes needed. SnapLogicโs integration with AWS IAM and Security were critical to building a compliant solution in a highly regulated environment.
The DCU effort is one of the first solutions to integrate SnapLogic with AWS Fargate - a serverless architecture that leverages the AWS Elastic Container Service. In addition to the rapid scalability, SnapLogics integration with multiple AWS services including Secrets Manager, Key Management Services (KMS) and encryption in the S3 Data Lake is unique. SnapLogic allows us to split out columns based upon sensitivity and place those columns in different RedShift schemas to enable appropriate user access to data with different levels of privacy (non-sensitive, sensitive and secure).
SnapLogic is heavily leveraged for data flows between systems (DB2/SQL Server/MySQL โ Parquet files in S3 โ Redshift). For operations within Redshift, we have SnapLogic generate dynamic DML based on metadata and โpush downโ data transformations to Redshift.
Using SnapLogic tasks that can be executed via URL calls, we leveraged an on-prem enterprise scheduler that allowed DCU to link internal dependencies to when SnapLogic tasks should be launched, and the timing was different and advanced taking into account banking holidays, as well as end-of-month timing differences. As opposed to just building APIโs with SnapLogic, we are using SnapLogic as an ELT / ETL tool to build out a full blown modern data platform in the cloud, including a Data Lake in S3, Parquet tables, and well-architected data models in Redshift and Aurora to support advanced BI tools like Tableau and soon to come advanced data analytics.
The dynamic approach taken with the pipelines means that we only need to develop a set of templates based upon what the source system database engine is. SnapLogicโs ability to generate code โon the flyโ, allows pulling in metadata and acquiring new tables, without any additional development necessary. SnapLogicโs ability to create complex execution paths within the pipelines allows for conditional execution of pipelines depending upon the nature of the source table (is it a type 2 or a type 1 or an append of transactional data).
The project currently has 23 FTE working on this solution. We have worked with SnapLogic in the integration with AWS native services and specific performance tuning of the PII data being protected. In particular we have engaged with the SnapLogic customer success team, which consisted of Praneal (sales) and Roger (engineer.)
The project is still in process with an anticipated completion within the next few months. To date, the project architecture, AWS Services and SnapLogic automation has achieved the following:
โข A well-integrated modern data platform in the cloud leveraging on-prem source systems, several AWS services (FarGate, RedShift, Aurora, S3, Secrets Manager) with, currently, an executive dashboard (Tableau) on top of well-architected data model.
โข Executive acceptance of the data being presented, with their request to leverage the dashboard in lieu of legacy reports for all of their reporting needs by next year.
โข Initiation of projects to retire legacy on prem reporting databases, which will save DCU significant money in maintenance and human resources
โข Projects being initiated to start advanced analytics and data science efforts.
The solution is still in construction, but over the next 1 to 2 years, we anticipate achieving the following goals:
โข 80% reduction in manual processes
โข 20% reduction in duplicated data across departments
โข Scalable solution to address 100% of performance constraints
โข 30% reduction in data quality issues and corresponding costs and customer issues.
These transformations are supporting the customer objectives of well-managed data (read: proper integration solutions) leading to consistent metrics, and an ever-learning, high achieving organization.
Early commentary from the senior leadership team has been extremely positive regarding the teaming of EXL and SnapLogic. They have been impressed with the level of rigor we bring with our solutions, the speed with which we can deploy and the quality and learning we are delivering together.
Through proper integration (i.e. using our Data Model Accelerator plus SnapLogic solution) combined with solid business rules we aim to ensure quality analytics through sound architecture and best practices (across DG, DQ Mgmt and Data Mgmt).
The foundational impact of EXLโs implementation of our Accelerator and SnapLogic technology is changing the game for DCU with 2 key themes โspeed & flexibilityโ. The successful teaming of EXL and SnapLogic is already producing innovation for both partners as well as achieving the customer goals and transforming their technology platform, building a data-driven culture and positioning the organization for sustainable success. We believe that this could not have been achieved without the cohesive teamwork enabled through the partnership.