Is Data Democratization Overhyped? | by Julius Cerniauskas | Sep, 2023

Data democratization is currently one of the leading ideas when discussing more effective usage of business data. Yet, the democratization of data, or lack thereof, is also named among the top challenges to deriving more value from using data in organizations across USA and Europe.

Thus, it is time to ask whether we have been barking up the wrong tree. Is the hype for data democratization justified? To answer these questions, we need to look at what obstacles businesses and other organizations face when attempting to democratize data and what success factors for democratization need to be addressed.

As is the case for many relatively novel concepts, the definition of data democratization has not stabilized yet and might be understood slightly differently by various users of the term. Often, the democratization of data will be defined as enabling end-users within the organization to access more data on their own. As we shall see, this broad definition is not enough to capture the essence of democratization.

Simply giving the employees access to more data will not make them actually use data more or get better results. The definition of data democratization must include not only superficial access to data but also actual, practical, and ongoing empowerment of the employees to constantly use data and improve their analytical skill set. Therefore, the term “self-service analytics” might grasp the essence of such an intra-organizational open data framework more accurately than “data democratization.”

The benefits of enabling more business users to access larger portions of organizational data have been recognized in theory and tested in the field.

Firstly, data democratization leads to more effective usage of time and resources. Citizen data scientists, defined as domain experts who are also capable of using and manipulating data proficiently, are crucial for a business to achieve such efficiency.

On the one hand, citizen data scientists take the mundane data management tasks off of the hands of professional data scientists, allowing the latter to concentrate on higher-level duties. On the other hand, domain experts can also do their job faster when, instead of requesting data-related services, they are able to fetch and work on data themselves.

Additionally, democratizing data brings it out of the dark. On average, globally, more than 50% of the data gathered during various company procedures is stored yet unutilized, taking up storage space, exaggerating data security and management costs, and burning electricity. Providing wider access to company data can significantly reduce the volume of dark data and create opportunities for extracting value from what is now only a burden.

Furthermore, when more teams with wide-ranging expertise are allowed to look at the data, better and well-rounded decisions are bound to be made. Individual bias and preconceptions that form due to particular professional backgrounds would have less of an effect, making data-based reasoning more versatile and objective.

Finally, access to data empowers employees and helps create a sense of ownership. Aside from the direct benefits of improved skill sets and operational speed, such empowerment also boosts the morale of the workers, who now have a deeper understanding of the big picture and their role within it.

Since every organization resembles a society that chooses its political order, we can understand the challenges of democratizing data and shifting to self-service analytics better by comparing them to the ones that obstruct the sustaining of democracy in nation-states.

In most democracies, all citizens have the right to access information held by governmental bodies. However, without established procedures, comprehensible cataloging, and digital infrastructure, few will see the records locked away in vast archives.

Similarly, just granting data access to more employees will not do much to expand the set of data users. Creating an effective data democracy requires not simply formal freedom but also substantial resources to use data. It includes providing employees with analytical tools that would allow them to work on internal data or even collect additional publicly available information useful to completing their tasks.

Implementing this in real life is complicated as the company must find or develop a custom solution that would have an SQL interface for experienced data users as well as some sort of “drag and drop” function for beginners. Such tools do exist, but they require well-documented data modeling. At Oxylabs, we chose Apache Superset, however, there is no “one size fits all” solution, and every organization must look for a tool that best fits their employee skills and needs.

It brings us to the second major challenge — that of literacy. Education is crucial to effective democracy of all kinds. Thus, proper training and opportunities for sharing practical, legal, and ethical data-related knowledge should be established. Without all this, there is a risk that employees will fail to use open data or, worse, interpret data in a wrong way, drawing incorrect conclusions.

Unfortunately, making sure that everyone knows the vast context of business data and uses the same definitions (such as what is “active user” or “product sale”) creates a lot of additional work for data teams — the primary challenge that data democratization should actually solve.

When people who previously lived under dictatorship transition to democracy, one of the main challenges is learning to take up individual accountability for public well-being that comes with freedom and self-governance.

In the case of the transition to data democracy, the same challenge arises in the form of accepting data protection as every employee’s shared responsibility. When managing and sharing data is the exclusive domain of IT and data specialists, the responsibility to ensure the quality, security, and compliance of these procedures also falls into their hands. This might be preferable to a considerable part of the workforce, as new forms of responsibility can be scary.

The goal, then, is to convince employees not to be afraid to use data and empower them to do it in a safe and responsible way. The aforementioned all-around training in data regulation and cybernetic security is crucial for achieving this goal. Airbnb showed a good example by setting up an internal Data University to improve data literacy among its employees; previously, the company opened internal data access by launching a data exploration tool called Dataportal.

Netflix and Uber, among others, also employ self-service analytics tools that are complemented by internal data training. For companies that decide to follow this road, the most important factor to consider is whether data democratization should involve all their employees in the same way (which is rarely the case). Often, organizations benefit from separating employees into different target groups or personas that have different data access rights and receive custom education.

In the case of democracy, a culture that motivates collaboration and decentralized decision-making is vital — failing to develop a fitting culture can render favorable material conditions ineffective. Even if all necessary tools are available and most employees are ready to take up data security responsibilities, democratization will fail without an organization-wide cultural shift. Such a shift underlies the following success factors for democratizing data.

• Top management support. The management agreeing to democratization initiatives does not necessarily mean that they are prepared to actively support it. Without top management oversight and demonstrated interest, sliding back into doing things the old way is all too easy.
• Having a clearly-defined vision and plan. Declaring democracy is one thing, and building it is another. When there is no vision regarding the role of open data in the organization and no action plan, developing data democracy rarely goes beyond the initial steps.
• Willingness to share data between different teams. Democracy is an open society based on mutual trust and collaboration. However, teams within an organization might be reluctant to share data with others, seeing it as a bothersome additional task or even stepping on each other’s toes.
• Promoting and enabling data-related initiatives. Data-driven initiatives by non-expert business users won’t necessarily produce the desired results or do it fast. Encouraging the workers to continue experimenting requires decisively choosing the potential of innovation over short-term productivity.

None of this can be achieved in a blink of an eye. Changing organizational culture is a long and bumpy process with plenty of opportunities to either get discouraged or rise to the occasion.

As the phrase made famous by Winston Churchill states, democracy is the worst form of government except for all the others. The same goes for data democracy.

Attempts to democratize organizational data are bound to be flawed and face numerous obstacles. Yet, it is the best way forward that we know of. Keeping the ever-growing data resources locked away and most business users completely dependent on a few data experts is not a sustainable alternative.

The hype for data democracy as a miracle solution to every organizational problem dies down when faced with real-life obstacles. Then comes the time for serious collaborative work, cross-organizational sharing of experience, and accumulation of knowledge to develop conditions for more democratic management of business data.