What does it mean to become a data-driven enterprise?

February 11, 2016
By Victor Nilson

Not everyone was a true believer at AT&T when we started our internal “Big Data Center of Excellence” three years ago this month.

It was obvious that a large, complicated enterprise – with 100,000 servers of disparate business data – would benefit from better organizing the petabytes to create new insights. But the skeptics had a point.  “Big Data” has enjoyed quite a bit of hype.

The road to becoming a truly data-driven enterprise has potholes. But here’s the good news. We’re proving that a business can motor through three critical phases if it stays focused: Building a capability. Proving value. Scaling the benefits.

We’re between those second and third phases, having already achieved considerable advantages from our investment. I’m excited about what this means for better decision-making, continuous innovation, and particularly for better customer service.

The typical enterprise runs on applications. Applications produce data that tend to reside within the application. That’s old school. When we surveyed all the data we had across AT&T – and ultimately had to invent part of the tool to do that – we found 70 million columns of data. That’s columns, not rows. And this was the most interesting insight: Only a small percentage of those columns were actually unique.

Is that easy to rectify? No way. The same data might have multiple methods of time-stamping, depending on the application. It might have multiple ways of storing a name or date. After all, the data warehouses in every business – yours and mine – are built in association with unique applications.

That must change. The successful 21st century enterprise will be data-centric, not application-centric. It will gain new insights without struggling to “get the data,” and at the same time it will be able to secure and control its internal data ecosystem more effectively. This includes “privacy by design,” where projects are reviewed for privacy compliance before they’re even tested.

Let’s talk specifics. These are just a few of our data-informed projects, in various stages, from proof-of-concept to full production.

  • Online chats with customer service don’t produce rows and columns of data – just lots of words. Our Big Data analytic tools can make sense of all those customer voices, letting managers know on a daily basis what needs to be addressed.

  • The best way to fix an Internet problem is before it happens. We’re tracking 53 metrics on our landline network to discover historical clues that reveal when a cable is about to cause trouble. By using predictive analytics to send crews out proactively rather than reactively, a pilot program showed a 40 percent reduction in total repair dispatches.

  • And about those trucks. Wouldn’t it be nice to know when they were about to break down? We’re using telematics from the trucks, historical repair records, and even local weather data to predict when batteries are about to quit – before they do. We’ve gotten our model up to 84 percent accuracy. That means the AT&T Entertainment Group can get an installer to your house on time.

  • I’m most excited about the ways in which we’re working to personalize your customer experience. Each time you log into the myAT&T app, go to att.com or visit an AT&T store, the most helpful information should be awaiting you.
Offices of Big Data at AT&T

Becoming a data-driven enterprise is partly about choosing the right data tools, and it’s partly about having the right people. More than 13,000 employees have taken online and in-person Big Data courses offered by AT&T.

It’s mostly about the right mindset. Using Big Data techniques, on big amounts of data, creates more meaningful data sets for insights. The scale and complexity of an enterprise thus becomes an asset instead of a liability, as discovery and learning are enriched.

We’ll continue this conversation in the coming months.



Filter by:

  • Most Popular
  • Author
  • Tags

Archives

Link to archives