Data Governance in practice: what must work to avoid fatal failure

Disclaimer: This article was translated from the Czech language by AI.

As the era of generative artificial intelligence begins, Data Governance is becoming an increasingly high priority, as the quality and description of corporate data will directly determine the effectiveness with which AI will be able to use it.

In an interview with Ondrej Kožár, Head of Business Analysis, and Michal Machata, Head of BI Excellence at dolphin consulting, we tried to uncover what all needs to work within Data Governance in order for it to not just survive “on paper”, but to start delivering real value to companies on a sustainable basis.

How to understand the concept of Data Governance?

In practice, the term can be understood as a set of rules, responsibilities, tools and processes that define how an organization manages, shares, describes and protects its data. “It’s not just about technology, Data Governance is an approach – a systematic way of taking care of data across its entire lifecycle,” emphasizes Michal Machata.

Ondrej Kožár adds, “It’s a cross-cutting topic that connects IT, business and management. And it is this interdisciplinary nature that makes it such a complex discipline to implement.”

Don’t start by buying a tool. Start with a strategy.

One of the most common mistakes made when implementing Data Governance is the assumption that choosing the right tool is the key to success. “We encounter an approach where a company buys a technology for metadata management or cataloguing and expects to have Data Governance covered – that the rest will ‘somehow’ fall into place,” says Ondrej Kozar. In reality, however, the technology tool is just a supporting tool. Without a clearly defined strategy, defined roles, established processes and active involvement of people across the organization, its potential remains untapped.

The first step should always be a data strategy. This defines not only the direction the organisation wants to take in terms of data, but also its ambitions, the architecture of the data environment and the level of centralisation or distribution. “Data Governance will look different in a company with a central data team and different in an organisation that has opted for a data mesh concept,” Michal points out.

The decisive factor is not the technology, but the way in which the responsibilities for the data and the processes used to manage it are defined. Without the right processes in place, the description of data in any tool will quickly become outdated and the tool will fall into oblivion.

Roles that it can’t do without

For Data Governance to work sustainably, roles and responsibilities must be properly defined.

The Chief Data Officer (CDO) tends to be the chief architect of the strategy and the driver of change. “If a company is serious about data governance, it can’t give this role to someone part-time. It’s a full-time job with overlap into change management, communications, marketing and process management. The CDO role cannot do without the proper authority.” says Michal Machata.

Furthermore, Data Owners are needed – i.e. people who are responsible for specific data domains. They must understand the business, have decision-making powers and be able to take responsibility for data quality. And finally, Data Stewards, who form the bridge between the technical and business worlds. “They have an overview of the data in the context of a particular system, understand its structure and logic, and often act as the first point of contact for data consumers,” adds Michal.

Motivation is one of the fundamental pillars

Setting up roles alone is not enough. If employees don’t have a reason to care about data, change won’t come. “It’s one thing to formally identify data owners. It’s another thing to make sure that the person has the motivation to fulfil their role,” says Ondrej.

One possible approach is to introduce KPIs tied to data quality and adherence to set processes – for example, for data domain owners. Another way is systematic education, explaining the benefits and ensuring that people have access to data in a form that they can work with it effectively. In some organisations, actively promoting data changes and their benefits – for example, through regular data newsletters or internal reports – has also proved successful. “The role of the Chief Data Officer is, among other things, to promote the benefits and business value of implemented changes in the area of data management – either personally or through the team that covers this agenda,” says Ondrej.

Processes as the backbone of operations

Data Governance without processes ends at one workshop. The sustainability of the entire initiative lies in a well-designed and followed process framework.

These include:

• Demand management – i.e. how the organization captures and processes data change requests.
• Data quality management processes – how errors are handled, who identifies them and how they are resolved.
• Release management – how awareness and consumption of data changes is ensured.

Selection of appropriate tools should be an integral part of this. Not only for cataloguing but also for process management. “The user should not be able to circumvent the rules. If I want a change in the data warehouse, there must be a tool that will not allow me to deploy it without the necessary documentation and approval,” says Michal.

Business dictionary and live content: what gives Data Governance concrete value

Data Governance is not just about managing processes or setting responsibilities. Its benefits only become fully apparent when real content starts to form – a business dictionary, a data catalogue, a catalogue of reports.

The business dictionary in particular plays a key role in unifying terminology and meanings across the organization. “Terms like margin or product can have fundamentally different meanings in different departments. Without a company being clear about what each term actually means, each analysis will be based on a different foundation,” points out Ondrej Kožár. Formally, this means not only definitions, but also calculation rules, domain context and responsibility for their correctness.

Experience shows that most companies will at some stage hit the limits of their language inconsistency – for example, when discussing KPIs or making decisions across departments. “Ideally, every new business need, every change in reporting or in the data itself, should be based on the dictionary – or updated in time,” adds Kozar.

The problem tends to be the sustainability of these descriptions. “We encounter that a company creates definitions, but a year later they are out of date – processes have changed, exceptions have been added, but no one has reflected the changes back into the dictionary,” says Michal Machata. The solution is to firmly embed updates into the change lifecycle – for example, linking them to business analysis processes or regularly reviewing them as part of a quarterly audit. A maintained dictionary and catalogue are not just a documentation obligation. They are a prerequisite for data to be interpreted uniformly, consistently and with confidence across the company. And this is how Data Governance becomes a truly functional framework for working with data.

Where will Data Governance go from here?

The final question of the interview was about the future outlook. Both experts agree that although Data Governance is not evolving by leaps and bounds, companies are beginning to understand its importance.

“Particularly with the advent of AI, it is becoming clear that without good quality, well described and accessible data, it will be impossible to realise greater ambitions. This is what can finally bring Data Governance to the centre of management’s attention,” concludes Michal Machata.

“It’s a long run, but those who go about it thoughtfully, with an emphasis on people, processes and strategy, have a chance to transform their data infrastructure from a passive repository into an active decision-making tool,” adds Ondrej Kožár.

Author: Jakub Holubec, dolphin consulting