Image credit www.iqlect.com/
Teradata CTO Stephen Brobst opened his presentation at Gartner’s Data & Analytics Summit in Sydney this week with a dire warning: companies must fully exploit data to survive.
“We have a saying in Silicon Valley: in the 21st century all companies are data companies, will be data companies, or will be extinct. Those are the only three choices.”
That seems somewhat extreme, but Brobst said it was rapidly becoming a reality in some industries, telecommunications in particular.
“Telcos think of themselves as providing voice and data services and so on, but that model is not going to succeed in the future. They are going to have to move to a different model where they monetise their data and where the value of the data is going to be higher than the value someone would be willing to pay for a telephone call and data services.”
He said this scenario was already playing out in India “where Jio is now giving away free voice services and data services that are basically 20 percent of the market price in exchange for data and then figuring out how to monetise the data.”
Jio decimates Indian mobile competition
I asked him after his presentation what had been the result of this initiative and he told me it had decimated the competition. “In 2018 Jio got more than 80 percent of all new SIM cards. They are pushing other people out of the marketplace. Vodafone ran away because they could not compete. Telenor ran away.”
It’s not known how Jio is extracting from this data value equivalent to revenue forgone. Brobst said that, as just one arm of India’s giant Reliance conglomerate, Jio was making use of the data only internally, not selling it to third parties, and has not disclosed any details.
However he did offer one source of insights. “One of the co-authors of a book called Sentient Enterprise is on the [Jio] board. His name is Mohan Sawhney. He’s a professor at Northwestern University and the Kellogg School of Management.”
The Sentient Enterprise: The Evolution of Business Decision Making, according to Wired “outlines [its authors’] five-stage approach to building what they call the ‘autonomous decisioning platform,’ a strategy that has been created through research and interviews with top executives from over a dozen blue-chip organizations including Dell, Verizon, GM and Wells Fargo.”
Wired said: “Using this five-step process [the authors] chart a course for readers that ultimately arrives at a new model for analytic capability, maturity and agility at scale.”
What Brobst did not mention was that the other co-author Oliver Ratzesberger was at the time COO at Teradata. (He became CEO in January 2019 but departed in November “by mutual agreement”. Teradata’s stock immediately lost 18 percent of its value).
Jio might be an extreme, and not easily replicated, example of the disruptive power of data analytics, but where it has led, others will in time follow or, if Brobst’s predictions are correct, fall by the wayside.
Data scientists at odds with IT
Trouble is, according to Brobst, organisations are being slow to exploit the power of data analytics because of the difficulties, and reluctance, to translate deas into action.
“One of the big issues is the culture clash between productising data versus innovating with data,” Brobst told the Gartner conference. “What Gartner calls mode one and mode two. Mode one people are trying to deploy in production, commit to service levels, scale up, manage costs down, and they are very risk averse.
“Mode two people need to move fast. They’re not thinking about scaling up, they just need to prove the value. They’re looking for the good ideas, and they are behaving very differently.
“The mode one people are the traditional IT people, and the mode two people are the traditional data scientists. … They come from very different worlds and you have a culture clash. You need an architecture that supports both productisation of the data, as well as innovation with the data.”
Trouble is, such an architecture seems to be sadly lacking. Brobst later told me that more than 50 percent of the models created by these smart data scientists never see the light of day, never go into production.
“That’s not because they’re not good, it’s because of the disconnect between the IT people who don’t want to support them,” he said. “Data scientists are not engineers. They don’t build things in ways that are sustainable and scalable.”
Is Agile the answer?
The answer, Brobst said, is to use this Agile DevOps approach rather than the waterfall approach, and while Agile and DevOps are now well established in software development, he said these techniques do not translate easily to building robust data analytics tools.
“A lot of the learnings are software driven, and there’s not enough people that know how to do it from a datacentric point of view. We call it analytic ops just to make very clear that agile analytics is not the same as agile software, because analytics is not data.
“Data has a very different gravity, and the kinds of trouble, the pitfalls that you run into in a datacentric world are different from the pitfalls you encounter in a software centric world.”
So it seems a lot of potentially very valuable data analytics initiatives never see the light of day. There will always be barriers to realising full production versions, and resource limitations, but if survival depends on ‘becoming a data company’ it will be essential that the business potential of all data initiatives are carefully assessed, those judged most valuable given priority, and the “mode two” analytics people retained rather than leaving, disillusioned by lack of uptake of their ideas.
