Data warehouse or data-lake or lake-house is a core component of your data analytics program. (In this article, I am using the term data warehouse in a generic sense) It pulls together data from internal applications/data-sources such as core banking, lending apps, origination, collections, marketing, finance, etc. as well as from second and third party apps into a single repository for critical analytics support decision-making.

 Here is a typical high level architecture of a data warehouse.

                                                                                                   Source - https://www.ibm.com/cloud/learn/data-warehouse

A typical data warehouse requires a database, a data migration or ETL or ELT tool, and a business intelligence tool to build reports and dashboards. When looking to get a data warehouse, one big decision needs to be made early on, which is whether you want to create a data warehouse or buy one.

Best-of-breed Solution

When you want to build your data warehouse, you get best-of-breed technologies and architect, design, and develop independently. You have total flexibility regarding what data you want to bring to the data warehouse, how you organize it, what reports and dashboards you want to build, or what AI/ML models your business wants. Of course, to do all this, you need a talented team. e.g. of best of breed technologies include Snowflake, Tableau, PowerBI, Talend, Qlik, etc.

Full Service Solution

These days, some vendors have credit union industry-specific data warehouses. Let's call them full-service data warehouse solutions to differentiate them from the best-of-breed data warehouse. These solutions include a preselected database, a data migration tool, and a business intelligence tool. To add value to their solution, they have built data ingestion from commonly seen applications such as core banking, loan origination systems, and digital banking. Etc. They continue adding more data sources to their solutions, from where they can seamlessly ingest data. The solutions also include industry-specific schema and pre-built data migration and transformation pipelines. Some solutions have added a bunch of data quality checks and industry-specific business rules to these data pipelines to add value to their solution further. These solutions include dozens of out-of-box reports, dashboards, and AI/ML-based models. e.g. of full-service solutions include Aunalytics, Tibco/ADMS, Arkatecture, Trellance/M360, VeriCU, etc. to name a few.

Comparison 

Both these approaches are very distinct and have their pros and cons.

The best-of-breed solution offers you maximum flexibility but requires significant effort up front in procuring the right set of technologies, installing them, and then administrating and operating them. All along, you need to have a reasonably good understanding of the best practices, define and document the design and standards, set up processes, etc. Significant advance thinking and work goes behind setting up a data management solution using best-of-breed technologies. When organizations select best-of-breed technologies and start the development with little or no design and standards for their architecture, design, development, operations, etc. then may see some early results but are very likely to have challenges in scaling the program from a long-term perspective. Any flawed design decision made in the process is likely to have significant long-term impact on the overall program.

In contrast, full-service solutions give your data analytics program a jumpstart with a preselected database, data migration, and business intelligence. All these technologies are integrated and work well together. However, these solutions do come at a cost. Also, you are now dependent on your solution provider roadmap for solution capabilities that may be critical to you. e.g., The ability to ingest data from an application the solution does not ingest from today or a transformation, business rules, or reports/dashboards that may be critical for your business.

When organizations select any solution without well-defined requirements then they are likely to miss critical success factors out as part of the evaluation process. This may have a substantial impact on the value that the program is expected to deliver as there are exorbitant switching costs.

These two approaches can be seen as two ends of a spectrum, with quite some options for you in the middle. You can decide to go either way or take a hybrid approach. Some vendors also allow you to take a full-service approach to jumpstart your data analytics program and then let you build from there. In that situation, you are essentially committing to their preselected data warehouse technologies, data warehouse schema, development standards for data pipelines, and business intelligence.

The decision to come up with the right approach to build your data warehouse is very personal and strategic. It is very much tied to your credit union's vision, strategy, scale, and data analytics talent, among other things. One thing for sure is that whatever decision you make has some long-term implications for the actual cost, opportunity cost, data analytics team, and overall success of the data analytics program at your credit union. Taking an incorrect decision may lead to a higher overall cost of the solution, longer timelines to get value from the investment, unsatisfied stakeholders, and missed expectations. Therefore, this decision should be taken with appropriate caution.