Our Hybrid Kimball & OBT Data Modeling Approach

Raw data is exactly what it sounds like. It’s unfiltered and difficult to parse for insights. Many businesses have a lot of raw data, but they don’t know how to refine it and unify it with other data sources so they can use it to guide their decision-making. That’s where Brooklyn Data comes in.

We help you turn your raw data into valuable insights that drive business impact with data modeling. We use a hybrid of two types of data modeling to transform your data, so it’s streamlined and actionable: Kimball’s Dimensional Modeling and One Big Table (OBT). We’ll discuss both methods by sharing how they work and why we rely on them together, but first, we’ll define data modeling and explain why we do it.

What is Data Modeling and Why Do We Model Data?

Data modeling creates an organized and structured representation of a business’s data to unlock insights and enable decision-making. However, it goes beyond simply creating tables in your data warehouse; it's about revealing the connections between data and enriching raw data with business logic to uncover deeper insights and provide greater meaning. We model data so that we can create reliable and accessible datasets to answer business questions and build standardized reporting. Without carefully modeling data, your organization can end up with fragmented data tables that are difficult to glean meaning from.

What is Kimball’s Dimensional Modeling? What is OBT?

Kimball’s Dimensional Modeling and One Big Table (OBT) are two popular data modeling approaches that we use. Before we explain how they can work together, we’ll share specifics about each one.

Kimball’s Dimensional Modeling

Kimball’s Dimensional Modeling, aka “star schema” translates business processes into two types of tables called facts and dimensions. Fact tables store business facts, typically quantitative business metrics, while "dimension” tables provide descriptive attributes of facts, giving them meaning. Kimball’s method organizes data around specific business processes, rather than by data source or functional department. Fact tables are defined at the lowest applicable level of detail, ensuring the highest flexibility during analysis, and allowing aggregation without losing any information.

One Big Table Modeling (OBT)

OBT data modeling elects to model data in — you guessed it — one big table. This means OBT includes dimension data in the same table as fact data, creating a single, wide, denormalized table containing all the data you need for analysis. This can be a huge table to store, but it’s efficient for computation since the table contains much information that’s aggregated and materialized in advance. So BI tools can perform aggregations and visualizations with significantly fewer table joins. Modern cloud data warehouses have incredibly cheap storage rates and charge more for compute, which has contributed to the rise in popularity of OBT data modeling.

The caveat is that while OBT can greatly improve query performance, it can increase complexity as there is a higher chance of data redundancy across different OBTs. Also, as OBTs get larger and wider (and they always do), cost implications may come back into play. This is particularly true if the joins needed to materialize the OBT are inefficient, and incrementality is not well-thought-out.

Why Do We Default to a Hybrid Kimball/OBT Methodology?

It’s Simple and Business-Driven

What we love about Kimball’s method is that it’s relatively simple. According to The Data Warehouse Toolkit, “Practitioners and pundits alike have recognized that the presentation of data must be grounded in simplicity if it is to stand any chance of success. Simplicity is the fundamental key that allows users to easily understand databases and software to efficiently navigate databases.”

Kimball’s method is adaptable to various industries and business models and more widely applicable than other data modeling methods, such as DataVault and Activity Schema. Its business-driven perspective allows end users to analyze the most important parts of their business, which are in the fact and dimension tables created. Kimball’s method is also flexible to changing business needs and easy to iterate upon. By using a dimensional model as the underlying structure of our data models, we set up a great foundation to build off these models for downstream use cases.

It’s Adaptable for Downstream Use Cases

To achieve our goal of using our data to enable decision-making and unlock insights, we want our data to be easily digestible by a BI tool such as Tableau or Sigma. We can achieve this by creating a “reporting layer” one step downstream from our fact and dimension tables. This reporting layer may look like a series of OBT data marts, which are differentiated by a department or business objective. This reporting layer allows us to change our level of detail from our fact table, add new relevant metrics, and create a central location for end users to query data relevant to their role. Separating the Kimball model from the reporting layer allows for cleaner tables that are more digestible for the end user, enabling smarter decision-making and insights.

Using Kimball’s method and OBT data modeling in tandem also provides a flexible structure for schema changes upstream. Our fact models act as a source of truth for quantitative business metrics, and as source data changes, we can update our facts and dimensions accordingly. Leveraging Kimball’s method before the reporting layer helps to break down the business data into core components, allowing bottom-up analysis, and slowly creating aggregated metrics as data flows downstream.

See How Our Kimball/OBT Hybrid Approach Can Help Your Business

Having a solid data foundation is a crucial investment in your company's future. Brooklyn Data brings the expertise and experience to implement data modeling best practices and turn raw data into valuable insights that drive business impact.

Interested in learning more if this hybrid approach of Kimball’s method and OBT data modeling is the right fit for your data needs? Reach out! We would be happy to share more about our process with you.