How a Universal Semantic Layer Enables Consistent Answers to Business Questions
The proliferation of data models and business logic across different tools has become problematic. The beauty of a universal semantic layer is that it supports many data sources and many BI systems and data applications so teams can continue to use their preferred tools.
- By Artyom Keydunov
- April 1, 2024
Estimates of the price corporations pay for using inconsistent or inaccurate data are staggering, as high as $3 trillion annually and costing an individual business 12 to 15 percent of its annual revenue according to The Harvard Business Review. In addition, silos and turf wars in organizations -- where different departments have competing data definitions -- erode employee productivity and well-being and cripple smart decision-making.
When Business Logic and Data Definitions Differ
Many examples of inconsistent data definitions causing disputes and distrust among teams come to mind. Sales and marketing may define customers and leads differently, for example, leaving management comparing apples to oranges with no clear idea how to achieve specific business targets. There’s also the case when separate teams with different business intelligence (BI) tools, techniques, and data definitions manage the data for their silo in different ways.
In the early ‘90s, BI providers sought to solve the problem and bring consistency to data through the concept of a semantic layer -- an abstraction layer between the source data and the output designed to offer a unified and consolidated view of organizational data.
To build a semantic layer, data teams first must create the business logic and metadata that goes into a semantic data model. They start by understanding the real-world questions the business needs to answer, gathering the necessary data, and then coding the relationships among the data (the business logic). Based on the business logic, they then create the semantic layer -- the abstraction layer -- using metadata that is defined so the data model gets enriched and becomes simple enough for the business user to understand.
Taming the Proliferation of Semantic Layers
The trouble is that different BI vendors -- from BusinessObjects to Tableau to Looker -- each introduced their own semantic layers over the years, and teams often chose different BI tools. In addition, departments often defined the business logic -- the relationships among the data -- differently. This has created a mess for data teams. They must constantly update, correct, edit, and add to business logic dispersed across the organization in a variety of semantic layers every time a new business question needs answering that involves different data definitions or business logic. This process generates massive duplication of effort, inconsistency, and manual effort with little business logic consistency and reuse, resulting in ongoing cross-functional disputes.
Enter the universal semantic layer. Separate from both the source and the output, the universal semantic layer is a “master” semantic layer that organizes, simplifies, and accelerates the consumption of unified data. With a universal semantic layer, everyone across the business must agree on a standard set of definitions for terms like “customer” and “lead,” as well as standard relationships among the data (standard business logic), so data teams can build one consistent semantic data model.
Consistency Without Sacrificing Preferred Tools
The beauty of a universal semantic layer is that it supports many data sources and many BI systems -- think of it as a single translation layer -- so teams can continue to use their preferred tools. Regardless of the BI tool users choose, the universal semantic layer allows them to work with the same semantics and underlying data layer, leading to insights and reports that are consistent and trusted.
Additionally, a universal semantic layer eliminates costly and redundant re-work across BI tools -- but most important, it builds trust within the business with a consistent, standardized “single source of truth” for data consumers.
How do you know if your organization needs a universal semantic layer? If you answer “yes” to any of the following questions, you do.
- Do your teams define metrics on their own, resulting in miscommunications? <./li>
- Do you have several lightweight semantic layers attached to various BI tools, causing disparate data experiences?<./li>
- Does your data team spend an inordinate amount of time updating and correcting business logic and metadata across different semantic layers?<./li>
- Do data teams have to create and update data in multiple places when a new business question needs to be answered involving data and business logic that are defined differently by each team? <./li>
- Do functional teams argue about basic metrics such as ROI or customer acquisition cost?
- Does your organization struggle with data reuse, governance, or auditing?
Benefits of a Universal Semantic Layer
These and other common issues point to the need for a universal semantic layer. A universal semantic layer is essential to leverage trusted, consistent data -- information that is correct, verified, and standardized organization-wide -- for better decision-making and less infighting about answers to business questions.
A universal semantic layer is required to power the next generation of data-driven businesses, accepting that there will be many different tools for visualizing and using that data and many different data sources where it is stored. Once an organization has built an intelligent “translator” for its data stack with its semantic layer and a single source of truth, there are many additional benefits. With standard data definitions, communication gets easier, fewer opportunities are missed, flawed decision-making is less likely, and arguments about answers to business questions become a thing of the past. The business goals leaders seek to achieve will be more easily within reach.
About the Author
Artyom Keydunov is co-founder and CEO of Cube, a venture-funded provider of a semantic layer for data apps. Prior to Cube, Keydunov co-founded Statsbot, a data platform.