How should an IT project manager drive a data governance project?

Data governance is often regarded as a "nice-to-have" project in organizations—it is an additional feature of data projects rather than a core component. This might be because it can't bring or can only bring limited practical value in the short and medium term. Whether this statement is correct remains to be debated, but we must explore this question: How can data governance be successfully implemented? Let's delve deeper.

Traditionally, data governance projects are implemented by the core support institutions of an enterprise's data strategy. The project is managed by the IT department and supervised by the data governance office. Although this approach achieved some results in the initial stage, it has problems such as lack of business autonomy, difficulty in scaling, and lack of continuous support. In the long run, it has lost its due relevance and primacy in the overall data strategy.

In the past decade or so, thanks to the progress of low-cost storage solutions, cloud computing, and artificial intelligence, enterprises have refocused on data governance, and its development model has shifted from centralized management to a more federated and decentralized approach. This shift stems from enterprises starting to build data products and opening up data to various use cases and consumption models. As the number of users grows, enterprises' demand for quickly and seamlessly accessing the right data through various access methods (edge devices, CDK, reporting tools, third - party computing applications) is increasing. This has prompted the rapid popularization of the data mesh architecture with a decentralized data publishing method, which is easy to understand, catalog, traceable, accurate, complete, and interoperable. Moreover, the ownership of this architecture belongs to the business department rather than the "IT department". Advanced data governance tools (such as AWS LakeFormation) that can automatically infer data lineage, generate data quality rules, and determine data classification (such as personal identity information) have further accelerated this shift.

Today, enterprises' awareness of data governance has reached its peak, and with the support of appropriate tools, technical project managers (TPMs) need to adjust their strategies to successfully deliver governance projects. Instead of running projects using a centralized approach, it is more reasonable to switch to a right - to - left approach. This means that business sponsors should play a central role in promoting data governance:

a. Collaborate with sponsors to understand key business initiatives;

b. Break down these initiatives into use cases;

c. Map use cases to published data products or identify new use cases when needed.

As a result, the focus naturally shifts to determining the data governance requirements of business initiatives. Therefore, data governance is no longer isolated but fits perfectly with the initiatives driven by business stakeholders.

Of course, some may argue that there is currently no overall plan to comprehensively manage all data attributes of an enterprise, and data governance is only carried out piecemeal. But this is precisely the key point—the "big - bang" approach to data governance doesn't work. Only by closely integrating data governance with business goals can it be effective! Moreover, enterprises can feel the benefits of data in the short term.

Over time, data governance improves, thereby achieving the enterprise's overall data strategy.

Here is an example of a feasible construction roadmap:

The business - oriented decentralized data governance approach, rather than the core model of the traditional data governance plan like the following.

In short, business value realization and data governance goals are frequently achieved in a short period rather than all at once as in the traditional approach.

This article is from the WeChat official account "Data - Driven Intelligence" (ID: Data_0101), author: Xiaoxiao. It is published by 36Kr with authorization.