Many companies are making significant investments in modern data platforms. However, there is still a key question: are these data architectures really efficient? And, above all, how can this efficiency be measured in data-driven environments? Without clear performance indicators, organizations often rely on rough estimates or expectations that are not always met in the actual use of data.
Faced with this question, the company Denodo proposes a measurement framework based on four key indicators that allow us to evaluate whether a data architecture contributes to generating revenue from data or if, on the contrary, it generates costs without being obvious.
Indicator 1: Time to Insight
The first indicator that Denodo proposes to evaluate the efficiency of a data architecture measures how much time elapses from when raw data is generated until that data is transformed into useful information for daily operations, strategic analyzes or data-based artificial intelligence applications.
When that process drags on, it’s often a sign that data is stored in silos or that the data architecture is fragmented. In these cases, you need to replicate data at scale, manually intervene in data processes, or add complex data integration layers. The consequences go beyond the technical field: when information derived from data takes a long time to arrive, decisions are delayed and companies lose competitiveness.
To measure this, it is advisable to calculate how long it takes for useful data-driven information to be available to the end user from the moment it is requested. This data “delivery time” can be analyzed in dashboards, analytical models and artificial intelligence applications, allowing improvements in data processes to be detected and results compared with other tools or during testing of new data platforms.
Indicator 2: Productivity in data engineering
The second KPI defined by Denodo focuses on one of the aspects that most influence the total cost of data platforms: the productivity of data engineering teams. When manual intervention or re-tuning of data processes is common, it is often due to fragmented data structures, duplicate data pipelines, or low reuse of existing data products.
This unmanaged fragmentation of data also has negative consequences on fundamental aspects such as data access control and data privacy. Elements that must be taken into account in order to put the data products being developed into production.
This indicator can be measured by the engineering hours required to integrate new data sources or develop new data products. It is also worth analyzing the relationship between data maintenance work and development work that provides value from the data.
Indicator 3: Infrastructure profitability
Inefficient data replication, time-consuming data processing efficiency losses, overstretched computing resources, or redundant data tool ecosystems can lead to unnecessary expenses. Therefore, the third KPI analyzes the real return of the data infrastructure.
To measure this indicator, Denodo recommends analyzing and calculating the total cost of ownership (TCO), including all relevant costs associated with data, such as computing capacity, data storage, data integrations, tool management and other factors related to the data ecosystem. Additionally, evaluating metrics such as cost per query, per use case, or per active user in data environments provides a clearer view of efficiency and scalability.
Indicator 4: Agility and adaptability
The fourth indicator identified by Denodo measures an organization’s ability to quickly adapt to new business requirements, market changes or regulatory demands related to data.
In the digital environment, this capability depends largely on the agility of data infrastructures, and implies the possibility of reusing data products in different data analyses, artificial intelligence applications or business initiatives. When deployment times for new data projects are too long, this is often a sign of rigid data management structures.
It is advisable to analyze the relationship between data maintenance work and development work that provides value from the data
An important aspect to be able to efficiently reuse data products is that both the metadata used as “raw material” and the data products themselves constitute a semantic layer of data that can be understandable by the different roles that intervene in the life cycle of the data product, including, of course, the consumers of the data.
This indicator can be measured by evaluating the time it takes a company to launch new data-driven projects, from the concept phase to their entry into production. It can also be analyzed how many use cases reuse shared data products and how many require independent data developments.
“Measuring all these key indicators is essential, but the true value appears when organizations can optimize them. In many cases, the problem is not the lack of data, but the complexity of the data architectures that make it difficult to access them quickly,” says Bernardo Godar, general director for Iberia & Latam at Denodo. “To achieve this, logical data management takes on an important role by introducing a virtual semantic layer between data sources and those who use them, allowing access to information without directly depending on where the data is stored.”
