There are several different ways to model your business and by consequence the data that supports it. Depending on your requirements are operational or analytical the model that suits them should be designed accordingly.

Typically, for operational requirements you go for OLTP (Online Transaction Processing) model and for analytical requirement the decision goes for an OLAP (Online Analytical Processing) model. The first it’s highly recommended to deal with your day-to-day business. High volume of transactions while preserving the Atomicity, Isolation, Consistency and Durability (ACID) of the same in the data store. The second, it’s much more focused to give your business additional value. That means, the data it’s organized and processed to answer business needs – either regulatory or strategic.

If your requirements are strictly analytic and the metrics well defined the most common OLAP model tends to be a traditional Star model (Star Schemas and OLPA Cubes). Although, if you pretend to create a model that represents the point of truth for further use – as for instance, by any Star Schema, you can consider to use a Data Vault approach (What is Data Vault?).

In a Data Vault the data that comes from your operational stream it’s stored without losing any of its original value allowing you to have a fine grade of detail of how each functional entity evolved during time. And mainly, because on top of this model it would be possible to build the metrics that business identify as long the need arrives.

The key concepts that differentiate the model are the following:

  • It’s a highly normalized model – specifically in the 3rd normal form. The functional entities are represented as hub tables, the relation between them in link tables and their characteristics in satellite tables.
  • The model allows to keep a track of the changes that occur over the entities, relations and characteristics. These changes track follows a Slowly Changing Dimension type (SCD2).

 

And because the model it’s all about functional entities and their relations, it’s highly extensible to address future business needs.

That said, the Data Vault seems to be a great way to centralize the data and made it available to your organization. But when it comes to design and implement a Data Vault solution there are a lot of concerns that should be taken into account.

At Link Redglue we help our clients to understand their needs in Data-Driven projects. Data modeling, either for analytical or operational models is between of area of expertise. To better understand if a Data Vault it’s the best solution for your strategy or either if you need help to put in place talk with us!

Samuel Ramos

Senior Data Engineer

#DATAARCHITECTURE #OLAP #OLTP #DW #STAR #DATAVAULT #DATAANALYTICS #NORMALIZATION #DATACENTRALIZATION