Click to learn more about author Tejasvi Addagada.
To read the first part of this series click here.
A well-defined Operating Model plays a vital role in achieving expected business benefits. Embarking on a Data Governance journey in an enterprise that spans divisions, geographies, and diverse stakeholders requires a good understanding of how various nuances of Data Management. As I start to define an Operating Model in an organization, I explore the choices of centralized and decentralized models from the perspectives of the benefits and readiness in the organization. If benefits arising from centralizing the Data Management operations across the enterprise is more than 20% to a de-centralized model, the focus should be to centralize.
On the flip side, friction in operations, reduced motivation in mid-level managers and bureaucracy at senior level often outweigh the benefits of a centralized model. That’s where the socio-cultural aspects have a major say in developing the Operating Model. Attaining a fine equilibrium between assessing, directing and managing data is essential. This equilibrium also enables self-service in the longer run, though there might be a need for initial re-skilling of personnel in distributed teams.
Again, I was working for a global bank, for its chief data office, based out of Bangalore. The bank headquartered in London has operations in more than 60 countries. As I started on this engagement, I got a deep dive into the existing Operating Model for Metadata.
I see the primary challenges as follows:
- Non-Integration between Data Quality, Metadata, Data Architecture, and other dimensions
- A federated Operating Model with stakeholders having difficulty accepting their responsibility to capture definitions and associated metadata
- A meta-model to manage Metadata that was not customized, based on the divisional needs
- Lagging timeboxed BCBS, financial crime compliance (FCC) programs
- Overemphasis on formalization of data programs through policy and rigidness in data operations
- Data Stewards hired, owners enforced with responsibilities but non-existence of differentiation between responsibility and accountability of activities.
A data definition along with other information helps us to understand the context in which the data element is being created or used. If two words I give, Mustafa and India come straight out of a data feed, can we derive the context around it? Below is the interpretation from someone to whom I asked this question. “Mustafa is the name of a person who lives in India.”
However, Mustafa is the name of a store which is a corporate customer to the bank, while India is the name of a district in Singapore. As we capture the definitions for the data elements, the context gets clearer. The actual name of the district is “Little India”, which got truncated to “India”. This truncation which is a data quality issue will be detailed in the next chapter. Metadata thus helps an organization to understand the context in which the data is being created or used. Without Metadata, the data is left for interpretation out of the context, and further implementation of changes to processes or systems will suffer as a result with inconsistent outcomes.
Here, I made it clear that, it is crucial for the success of a Data Governance Model to differentiate stakeholder responsibilities from accountabilities. Most of the Chief Data Offices fail to have this clear distinction in the way they operate. For example, a Data Steward can be responsible for defining thresholds for Data Quality while the data owner can have the accountability to define thresholds and metrics. That said, stewards are accountable for ensuring that the data owners accept their accountabilities.
Standardizing Metadata-as-a-Service will assist the organization in pushing or pulling the capabilities based on their needs. The data organization needs to emphasize promotion of existing Metadata capabilities. This is a stepping stone when the cultural change of “Managing Data as an Asset” is to be trickled into grass roots of the firm. The awareness sessions should be customized based on the nature of stakeholder groups. One wouldn’t want to pitch the word metadata to business users, operational and front end staff. As this often gives a technical fervor, I prefer to use “Business Terms” and “Associated characteristics” to refer to data elements and related Metadata attributes.
To have an organizational structure that supports future operational processes, certain activities need collective participation by a group of stakeholders. This primarily can include various functional levels and have representation from diverse divisions as well. One such stakeholder group can be a Data Stewardship Council, which can have a permanent representation from data stewards of divisions and geographies along with the representation from the Chief Data Office and other stakeholders like data owners on invitation.
Metadata Management also has a life cycle that needs to be aligned with Project Management, transformation and risk management lifecycle stages. But, is your firm actively managing the status of the Metadata across its lifecycle to manage the Metadata itself? There is a visual snapshot on the statistics of Metadata status across its lifecycle, viewed from the perspectives of domain, system, process, and user. For example, in Data Domain-A, 50% of data elements or business terms have been defined while only 30% are published.
The challenge for organizations has always been to harmonize disparate data across the organizations’ landscape. This is because data is referred to with different names, interpreted with different meaning, in a same division or firm. Classifying data into semantic, logical and physical models helps in deriving ownership at a high level.
Mapping relationships between the business terms and data elements in-fact simplifies the data landscape. It is a best practice to capture all the relationships associated with the data that will aid in better impact analysis, data analysis and requirement management. Every change in state of an entity like a “Lead to Customer” is associated with business rules like “A lead on completing a product purchase transaction is a customer”. Further, if there is a business rule that has multiple business terms, the business terms appearing in the rule can be related with a relationship like “Relates to” after importing them to Glossary.
There are immediate and cumulative benefits from actively managing and governing data. The focus of firms should be on differentiating Data Management from Data Governance to achieve the required cultural transformation. This includes people or stakeholders as major players. Further the focus is gradually shifting from having to mature Data Management capabilities to monetizing direct and indirect data benefits. It is strongly recommended that the Data Governance functions have an assessment plan before kick-starting standardized services be it data quality or any other dimension. For organizations that do not have a problem with Data Quality, governing their data will bring trust in data to the people leveraging this data for decisions.