One thing that drives analytics consultants absolutely batty is clients who believe their data is inherently clean, regardless. That’s on the business side, of course. IT, data science, and data engineers are all too familiar with what it takes to get data in order, but one thing even they may overlook is the usefulness of that data over the long term. Some are throwing everything into a data lake hoping for the best later, which is fine, as long as there’s some structure and governance in place.

GPU manufacturer NVIDIA is addressing that very problem as it attempts to enable self-service analytics. If the analytics are going to be reliable, then data quality and documentation need to be considered. As I write this, there’s a pilot project unfolding, and a big part of the effort is focused on the data itself.

“There’s not enough emphasis on data assets. It’s more like a byproduct of your systems,” said Ivan Chen, director of Enterprise Business Analytics at NVIDIA. “The first we’re doing is taking inventory of all our data assets, documenting them, and then trying to figure out how we can use that data effectively to make decisions and understand what it means.”

As part of the project, Chen and his team are documenting all the steps needed to meet a performance metric, so everyone can agree not only to the KPI but how that it is achieved.

“I want enable analysis, not give you a debatable report because neither of us really understands the nuances of the data,” said Chen. “If we define and document everything, then we can agree about what a data field is and how it’s used. That way, we can have a conversation about what to do with the data instead of debating about what wasn’t considered.”

The scope of the documentation includes memorializing the data transformation so people can understand how it was done now and later.

Modernizing the BI Platform

NVIDIA is in the process of modernizing its BI platform to enable self-service capabilities. Like other companies, NVIDIA has struggled to leverage pockets of data owned by different people, even though the data came from the same source, such as an ERP system.

“What we’re doing now is we’re bringing it all together in a central repository and documenting all that, so analysts can use it in a self-service way,” said Chen. “I’d like to scale our documentation that so more people understand what’s available and how they can use it, because we’ll be able to create more insights.”

The effort is an attempt to overcome what so many organizations have experienced which is business requirements that move and change at an Agile pace, and the Waterfall nature of IT creating reports, which is gathering requirements, building something, and then finding out whether it really meets the needs of the business or not.

“NVIDIA has a very progressive IT shop where they want to partner with and help business,” said Chen. “As part of the three-month trial, we’re testing the Agile method. There’s a dedicated team from IT that’s solely working on this so there is some sustainability.”

Part of working with IT is resolving the Agile-Waterfall disconnects because the point of self-service analytics is to enable faster and more timely insights.

“Analytics really needs to be treated differently than an implementation project,” said Chen. “I think IT has accommodated this because they recognize that Waterfall reporting is not going to work in today’s fast-changing environment. That’s why we’re able to try this new [Agile] method.”

If everything goes well, the project will scale up to the enterprise level eventually, which will enable more and different types of insights than are achievable today.

Achieving a holistic view is very important if you want to answer important cross-functional questions,” said Chen. “There are a lot of connection points that need to be understood and reconciled if you want to get to an accurate, holistic view. Most companies don’t have that yet.”