Data governance is a challenge for many organisations, especially when it comes to knowing what data is available and where it is stored.
Typically, we would implement a data catalogue to address this challenge. But getting the most out of these technologies is about more than just installing a tool. In this post, we want to share how applying the tools and techniques of a product manager can help to address data governance problems.
Typical data management challenges
We often work with organisations that have built up large and potentially very useful data estate. This type of organisation commonly experiences data management issues, including:
– High turnaround time to discover information around business and data assets
– Data analysts had low confidence in data due to limited shared context
– Lack of confidence in data sets because of a lack of lineage
– Siloed information leading to effort duplication
These issues cause enormous frustration. Business managers can’t see if projects were successful because it is hard to locate data, while data analysts are unsure if a given data set is trustworthy. Every department has its own definition of what a customer is (and their own data sets).
What does product thinking look like?
As product managers, we would apply a classic product-based approach to shape our response to these challenges. This is: discover -> prototype-> implement approach
Conducting user research helps to identify challenges the organisation has with information management, and can include workshops and individual meetings. These meetings should allow you to identify personas, pain points, frustrations and where improvements can be made. It’s also an opportunity to learn about strategic technical drivers, which helps to identify the most appropriate tools and technology.
By the end of the discovery stage, we had defined the bigger picture and associated vision, goals and nets. We also had a defined set of success measures, which were:
- Time saved by data analysts and other functions (either as a result of improved access or because it was quicker to find the right data asset.)
- Employee Empowerment – does a data catalogue improve employee knowledge of available data and does it make them feel more empowered?
- Increased speed to onboard new employees. Does a data catalogue improve how quickly you can onboard someone onto an analyst team?
- Reduced Infrastructure and related costs – does a catalogue enable people to find existing data sets or reports and does this lead to reduced infrastructure costs?
In a standard product approach, we will prototype the product to assess how people use it and evaluate whether benefits are likely to be met. Some organisations might not be convinced about the need for a prototype, but it’s essential to developing new products and introducing new services, and can make a critical contribution to success.
If it is important to get feedback on the catalogue quickly, it’s not necessary to implement a prototype of the whole service. Whilst automating metadata entry is a big accelerator for data catalogues, for the prototype we use handcrafted templates so that we can get feedback on the user experience quickly and understand what metadata was most relevant.
Once a prototype is in place, having one to one sessions with users helps gather feedback on their use of the tool. We look at how users are being helped to do typical tasks along with a post-test questionnaire that measures acceptance, usability and user satisfaction with the data catalogue. Some of the important points that will come out of this kind of evaluation might include:
- The catalogue displays various metadata for data assets. Using the prototype we can assess which ones are most useful
- Users can easily see where the data came from through a lineage function – this shows what the source is, and how close the asset is to the source. This really helps users assess their confidence in the data-set.
- It also shows a snapshot view – a small sample of the data, users should find this helpful in understanding the content.
During the prototype phase we could also learn:
- What is the most popular tool? Search, or something else?
- Do users find it helpful to request access through a basket function that automates the work of manually finding out who owns the data, making contact with them, waiting for them to have the bandwidth to provide data access etc?
- How useful are crowdsourcing tools? In one recent project users liked that they could provide ratings and reviews of data sources. This helped other users find what they wanted.
- What percentage of data sets are redundant, perhaps because they are duplicate or not useful?
This user feedback means that we are able to iterate on the prototype to improve the design. For example, in some instances, user feedback showed that users were overwhelmed by the standard metadata. In this case, we created customised views based on their feedback. Additionally, new users struggled to understand how to use the tool, so we created an introduction session that walked new users through how to use it.
Finally, the initial catalogue organisation was difficult to navigate so it was refined based on user feedback.
The actual implementation of the production catalogue was undertaken by the catalogue vendor’s implementation partner. In this phase, the supplier continued to give direction and worked with the client to measure the success factors.
We hope that this post has given some insight into some of the considerations when creating a data catalogue solution, and shows how product thinking isn’t just about user interfaces. It can also be applied to data to help organisations shape important data services.