The Hidden Potential of Data
Of all business assets, data offers the most hidden potential for virtually any organization. During their interview at GoDataFest 2021, Kevin Duisters, Commercial Analytics Lead at Interfood, and Eduardo Neves, CTO at Funda, discussed their experiences, while Xebia’s own specialists Bram Ochsendorf and Pádraic Slattery shared their insights as well. In this article, we’ll explore data democratization, the process through which data becomes available to a wide group of users within an organization.
It is important that data can be used for multiple aspects. Not only specialists but employees working in different departments should be granted access. Data democratization offers organizations the opportunity to be truly data-driven.
How Does Data Democratization Put Data to Good Use?
Data Democratization means that not only the engineers or data scientists can translate the data into concrete proposals. Instead, all users and decision-makers should be able to benefit from this valuable source. This is becoming increasingly important as data dependency is increasing. Only by putting the data to good use can the end-users make the right decisions quickly and based on accurate information, for example, to improve processes or to perform estimates more accurately. These analyses increase the understanding of customer behavior and stimulate the development of new services. Internal discussions can be driven more by facts and less by opinions or assumptions. As a result, meetings are more focused.
Improving Data and Quality Management
Data democratization is also a way to solve all kinds of existing problems, such as data availability or employees’ knowledge of how to use the data effectively. It offers an improvement in data and quality management. This ultimately increases the confidence managers will have in the data.
Previously, different departments were collecting the same data alongside each other and processing it, unnecessarily, separately. But now, data democratization can offer a streamlining of this process.
For example, Eduardo Neves, CTO of Funda, during an interview at Club Cloud 2021, said that his teams used to process more-or-less the same data separately six to seven times.
In other words, after successful data democratization, the use of data is no longer limited to the boundaries of teams or departments. The right information now gets to the right people at the right time.
“We had a function for the statistics for brokers and for the classification of offers based on user behavior, that’s based entirely on the same data. But because it was used differently, it was processed in a different way. Data was also used less than what was actually possible. We had to change our way of working”, Eduardo said.
Steps Towards Data Democratization
Data democratization has two components: the availability of data, and self-service analytics. Both are crucial. Making all data available haphazardly within an organization not only carries significant risk but is also completely useless if the organization cannot use the data effectively. Conversely, an organization full of excellently trained employees who know all about data can do little if no actual data is available. By investing in a modern data stack and data platform, for example, we can improve availability. Organizations are now starting to see data as a product, rather than a byproduct, backed up with good governance. Data literacy and support are important for the success of self-service.
However, the exact development of both domains varies from organization to organization. But in almost all cases it is done step by step. Interfood Group, one of the world’s largest dairy trading companies, for example, started by democratizing the insights of the Business Intelligence department.
“Then we started working upwards,” says Kevin Duisters, Commercial Analytics Lead at Interfood during Club Cloud 2021. “The final acceleration was provided by the data platform. It is the engine room of our data democratization, not only making technical resources centrally available, but also consolidating innovation.” Employees can use the platform to develop their ideas into prototypes and provide feedback on the results.”
The modernization of the data stack, with an enrichment using artificial intelligence (AI) followed, and then democratized that further.
When it comes to data literacy, many companies still experience hurdles. Research from Xebia shows that while 49 percent of organizations are using data for dashboards and reports, only 25 percent are developing predictive models. On top of that, 34 percent of organizations say they have difficulty building data application knowledge. This is in spite of the fact that 75 percent of all employees deal with data on a daily basis. Training is crucial so that all employees are at least aware of the importance of data. In addition, the IT, BI, and data engineering teams must be ready to provide support. However, their resources and time are limited. That is why it is ideal if a company manages to establish a community, through which experiences and best practices can be shared. All of this requires a high degree of collaboration.
“My advice would be: start with the data team,” says Duisters. “They have to be close to the business. We try to do this by creating a triangle of data scientists, data engineers and data analysts. They all have access to each other’s tooling, which shows that collaboration is definitely possible.”
Paths Towards Data Democratization
Tackles data democratization by guiding the process along three paths:
Path 1: Organize - This includes identifying where the organization currently is, where it wants to go, and determining the strategy.
Path 2: Set Up - Implement the data platform based on an appropriate cloud infrastructure and introduce an Analytics Engineer, a new role for the data team.
Path 3: Train - The search for the right balance between people and skills, and the increase of data literacy.
Three Risks of Data Democratization
Data democratization does not happen automatically. And in addition, it cannot be underestimated how meticulously the process must be carried out. This starts with the fact that simply sharing data is not permitted. Personal information can’t just be sent around internally as this can increase the risk of data breaches. Laws and regulations such as the General Data Protection Regulation (GDPR or AVG in Dutch) require strict compliance.
Also, a lack of data literacy can lead to the wrong conclusions being drawn from the data. A problem with this is that it is not always evident that the knowledge is inadequate. Employees may think they have it down after a training session, but it may also be that they still don’t understand something well or have been instructed incorrectly. A particular value can mean something very different on its own than it does when combined with other values. Therefore, it is crucial that employees learn to walk before they run.
The third risk of ill-considered data democratization is that different business units process and document data in different ways. This can lead to chaos and unnecessary extra work.
The Analytics Engineer
To cope with all that, a bridge is needed between data engineers and data analysts. To that end, a new role has emerged in data-driven organizations: the analytics engineer. Their job, made possible by modern data stacks, is to pull the best practices from the software world into analytics to create high-quality datasets. However, these sets need to remain somewhat flexible, which means they are not one size fits all. The whole idea of data democratization is that users themselves have the ability to do new things with data. That is why a good analytics engineer must know SQL inside out, and preferably know the data pipeline in detail. This is how they can take the data from IT and translate it to business.
Reaping the Benefits of Data Democratization
If the process is well managed and documented, data democratization can go a long way toward creating a truly data-driven organization. The ability to allow users to form new insights on their own can offer the organization a lot. And this allows data analysts to focus more on big projects within the organization.