Data risks become business risks when inconsistent numbers reach decision-makers, sensitive data leaks to unauthorized users, or metric changes fail to propagate to all reports and tools. The article presents the semantic layer as a practical way to mitigate these risks by placing metric definitions, business rules, and permissions in a central location that different tools can reference.
The discussion does not address risks in the abstract sense associated with compliance frameworks; rather, it focuses on the daily operational losses that may not appear immediately. A regulatory audit might reveal a discrepancy in a metric between systems, a board member might notice two conflicting revenue figures in two consecutive reports, or an artificial intelligence tool might produce a recommendation based on data that has not been governed since the departure of the analyst who created it.
Three Main Areas of Risk
According to the article, data risks are often concentrated in three interconnected areas, and organizations may face them simultaneously.
- Accuracy: As organizations expand their use of tools, dashboards, and artificial intelligence-enabled applications, the scope for error also grows. A revenue metric might be defined one way in a Tableau workbook, another way in a Power BI model, and a third way in a Python notebook. This difference is not merely a technical inconvenience; it may lead to poor strategic decisions, misallocation of resources, failure to meet targets, and declining trust in the data team.
- Governance and access: Access controls are typically distributed among the data warehouse, business intelligence platforms, dashboards, cloud storage engines, and shared drives. Each system has a different permissions model, management interface, and auditing capabilities, producing a fragmented environment that is difficult to maintain or review with confidence. Sensitive data may reach a dashboard where it should not appear, not necessarily because of malicious behavior, but because the governance surface has become too broad to manage consistently.
- Change management: A chief financial officer might decide, for example, to exclude trial customers from the annual recurring revenue metric beginning next quarter. But implementing the change may require modifying a view in the data warehouse, two Tableau workbooks, a Power BI model, an Excel report managed by the financial planning and analysis team, and an analytics tool that depends directly on the data lake. If some of these elements are not updated, the numbers will diverge again over time.
These risks intensify when a metric is not governed, is defined in different ways, and cannot be updated from a single location. The problem, therefore, is not necessarily that the change itself is incorrect, but that it may not be implemented fully across every part of the environment.
Limitations of the Traditional Model
Organizations typically rely on a centralized team of business intelligence analysts to serve as the gateway to metrics, reports, and dashboards. A user requests a new report, a change to a metric, or an explanation for a discrepancy between two figures, and then waits for the request to be processed. This model arose partly because organizations do not trust their data sufficiently to allow self-service, but it has clear costs: slowness, request bottlenecks, hiring costs, and inconsistent result quality depending on the analyst and the tools they use.
The problem grows with the distribution of access controls, quality reports, data lineage tracking, and business ownership across multiple tools and systems. Every additional dashboard, data source, or platform adds a new governance surface, another place where rules can differ, and a potential point of failure. As a result, data teams may spend more time on maintenance and correction than on delivering data and insights.
One Layer for Definition and Governance
The article proposes a different model: placing metric definitions, business logic, and calculations in one location within the semantic layer. If the annual recurring revenue metric is defined once, Tableau, Power BI, Excel, Python, and the artificial intelligence-powered conversational tool can all rely on it. When the definition is changed, such as by excluding trial customers, the change flows to downstream tools instead of requiring a manual search for every logical copy of the metric.
The article also connects the semantic layer with version management, enabling teams to track versions of key metrics and determine how a metric was calculated at an earlier point in time. In parallel, the semantic layer can become the central access point for governed data, allowing teams to use their preferred tools while permissions, definitions, and business rules are managed from one location. In this way, the scope of governance contracts from dozens of systems to a more focused point.
The proposed benefit is not limited to consistency. The semantic layer can carry field descriptions, metric definitions, relationship maps, and business rules alongside the models, columns, and metrics themselves. When data context becomes documented where the data exists, it no longer depends entirely on analysts’ memories or scattered documents that may not be updated. This supports self-service and also enables artificial intelligence agents to read the context and understand data more broadly.
Reducing Risks, Not Eliminating Them
The article emphasizes that the semantic layer is not a complete remedy. The underlying data still needs to be clean, organized, and maintained, and the rule “garbage in, garbage out” still applies. Standardizing metric definitions also requires organizational agreement and leadership commitment, which software cannot replace.
However, according to this argument, the semantic layer changes the economics of risk management. Instead of adding more employees and governance tools to keep pace with every source, dashboard, and platform, it reduces the number of places where logic can drift, changes can be lost, or auditing can become difficult. Risk management becomes more containable because it is concentrated in one location rather than distributed across the entire data stack.
This becomes increasingly important for organizations that rely on artificial intelligence-enabled analytics, since these tools need governed data accompanied by context in order to produce outputs that can be trusted. From the article’s perspective, the semantic layer is not merely an architectural improvement for consistency, but part of the infrastructure required to manage data risks in an environment where the cost of decisions based on unreliable information is accelerating.