Login
Register
How do you even begin to transform your data model? Telefónica's answer is to start with a set of core values and objectives that any useful data model must accomplish. Daniel Vaughan, Head of Data Science and Chief Data Officer, Telefónica Mexico discussed further.
How Telefónica’s data modelling principles shaped its global approach
Communications service providers (CSPs) arguably have access to the most comprehensive insights into customers’ usage and behaviors, leaving them well positioned to solve customers’ issues, fast, develop responsive new products and services, and enable digital ecosystems for emergent technologies such as the internet of things, artificial intelligence and blockchain. This article based on a presentation at Digital Transformation World 2019 by Daniel Vaughan, Head of Data Science and Chief Data Officer, Telefónica Mexico, sheds light on how operators can transform their data systems.
“The sexy part of data is artificial intelligence, machine learning, how we make intelligent decisions,” Vaughan says. “But unfortunately, you can’t be sexy without the right data. So, we have to create the right data model first.”
Indeed, research by McKinsey & Company shows that operators that get this right “get ahead of customer requirements and drive a 30%-50% cash-flow improvement through revenue acceleration and cost optimization.” So, how do telcos go about determining and implementing the right data model? Vaughan has some ideas. He is responsible for leading all big data and data science initiatives, heading the Advanced Analytics Center of Excellence and developing Telefónica’s data strategy. At Digital Transformation World in May, he explained how the principles behind the company’s unified reference model (URM) helped to form the foundations for its now infamous 2017 AURA project – the app that allows customers to control the data generated by using the operator’s products and services transparently and securely, and also the foundations for how it handles siloed data and data solutions in its local businesses countrywide. “I won't answer the question of the right data model for telcos in general,” Vaughan says. “Let me try to tackle a simpler question: What was the what's the right data model for our company?” The simple conclusion Telefónica came to is, “the data model that feels right for us.” But of course, nothing is that simple, and to find the right model the company had to ask some introspective questions: The answer is a set of core values and objectives that any useful data model must accomplish.
1 – Empowerment: The data Telefónica holds belongs to customers. The company does not own their data, and it must empower them to access and make decisions about their data at any point in time. 2 – Transparency: The company needs to be absolutely transparent about what type of data it has about its customers, and how it is being used. 3 – Security: Telefónica must guarantee at all times that customers’ data is safe.
“Data is a first-class citizen for Telefónica,” Vaughan affirms. “What that means is that it is a strategic asset for our company.” He explains that the three traditional strategic assets for telcos are: Their networks, systems and the commercial offering (products and services).
“The fourth strategic asset now is data, and as a strategic asset, it must be creating value in the short, medium and long term.”
Telefónica measured this using a fundamental proof of value – use cases. “We deploy use cases and each use case solves a very specific business problem. We create these use cases by making small decisions [for these business problems] using prescriptive and predictive modelling, machine learning, artificial intelligence – hopefully creating value.”
Needless to say, in today’s digitally fast-paced market, telcos need to move as quickly, if not faster than their competitors – who may or may not be part of the telco ecosystem. So, data models must deliver agility.
“But this sounds easier than it actually is,” Vaughan says. “Why? Because data models are like straitjackets: They’re really restricting us on the type of choices we can make in the future.”
He recommends CSPs think in terms of two concepts:
Telefónica found that its local operations were solving the same problems over and over. Each had developed their own solutions for network optimization, fraud detection and churn prediction modelling, for example.
“[There was] too much redundancy and geographical silos,” Vaughan says. “We were trying to break these silos and create a global knowledge base.”
Whatever data model Telefónica designed needed to provide structure, but as Vaughan notes, data models are constricting, and there needs to be some flexibility to be able to respond to market needs and changes, and subsequently create value in the short, medium and long term. On the flipside, too much flexibility may incur technical debt from quick, short-term solutions, debt that firms will inevitably end up paying in the future, reducing the long-term value they create.
It’s clear there is increasing pressure on CSPs to become data-driven so that they can deliver remarkable customer experience and innovate, particularly since the approaching surge of IoT connected devices will exponentially increase the amount of data generated. In a recent interview with Dr. Anastassia Lauterbach, Author & Entrepreneur, NED, and member of TM Forum’s Digital Ecosystems Advisory Board, she perfectly echoes Vaughan’s sentiments in her advice to firms on the importance of data as telcos develop their ecosystem strategies:
“They need to systematically review what they want and can do with data, where this data is trapped in terms of IT silos, how to enable data cleansing and processing to build models. This will all enhance top and bottom line. They need to understand centralized versus decentralized AI approaches. They need to develop an ecosystem which is right for them, as opposed to copying the approach from top internet brands.”
Interested in this topic? Download our recent report Transform service activation for 5G services with real-time data analytics
