Do you need an SAP data mart, or an SAP data warehouse?

There are so many data management architecture buzzwords these days: old school data warehouses, data marts, data lakes, logical data warehouses, and the list goes on.

In brief, a data warehouse is a very disciplined approach to managing data across an enterprise. Data warehousing attempts to establish a canonical version of an enterprise's data across business areas, data sources and time. This allows for powerful comparisons across the business in terms of an agreed-upon "'version of the truth," as well as comparisons between past and current performance. The cost is that this type of data management is labor-intensive, not particularly agile and usually bound to a particular point of view that has been agreed on by a critical mass of stakeholders across a company.

A data mart is usually a more focused implementation. Like a data warehouse, a data mart often uses data from multiple sources and spans a large time period, but it tends to be developed in the service of a particular business problem. Because of this focus, a data mart is often able to be better and more quickly adapted to addressing that particular issue than an enterprise-wide data warehouse. The downside is that different data marts will tend to give different answers to questions that appear to be the same, which can cause confusion within a business. This happens mostly because different data marts will define metrics slightly differently, or from different points of view, even if those metrics have the same name. For example, a tax data mart will often view the business as a collection of legal entities, each with their own revenue, cost of goods and overhead expenses. But a management data mart may view the business as a collection of functions. The answer to the question, "Where is value created in the business?" may be very different depending on which data mart one is looking at.

So do you need a data mart or a data warehouse? If you're asking the question, you should probably start with a data mart or something similar in scope. But start to research and understand data warehousing concepts. You'll soon begin to recognize problems that data warehouses are designed to address, and you'll then be able to make a more informed decision regarding whether it's time to implement a more comprehensive data management strategy, or if you will continue to get the most bang for your buck with smaller, more focused data marts.