The data explosion has created a conundrum for many businesses, which have bought into the big-data dream only to find that they still can’t unlock the value of their data.
Business units are intrinsically self-preserving, after all – and when that instinct drives line-of-business managers to hoard data they see as valuable, the efficacy of data analytics suffers.
Renee Lahti knows this all too well: as chief information officer of data-management consultancy Hitachi Vantara, she has toppled one conceptual wall after another while transforming the way the company saw and used its data.
Like most companies, the organisation initially followed core big-data tenets such as the creation of a ‘data lake’ where all of the company’s data would be dumped and, theory goes, available across the business to instantly provide amazing new insights.
The truth, Lahti found, was quite different.
“This whole ‘build it and they will come’ theory, where you dump all your data and every line of business will go find their value, didn’t work,” she said.
“The project went live and the CFO went looking for the return on her investment and it didn’t show up.”
Drawing on DevOps for data
The challenges of big-data ROI have been well documented, with Gartner predicting years ago that 60 per cent of big-data projects would be abandoned after failing to deliver.
Less than a year ago, the firm noted, 87 per cent of companies still have low business intelligence and analytics maturity.
The ongoing struggles of big-data investments have sent companies back to the drawing board as they realise all the optimism in the world isn’t going to deliver ROI without help.
ANZ chief data officer Emma Gray told the recent Melbourne Business School Business Analytics Conference that “the fear of doing the wrong thing has meant people have locked their data behind Fort Knox doors and made it very hard to access it.”
Working to outline a more coherent data strategy, Gray talked with business leaders who begged her not to, for example, share proprietary risk data with other business units.
“They were terrified to go have a conversation,” Gray said, “and the big fear was that I was going to take all that data and let [other business units] go to town and start doing really stupid, crazy things.”
Gray found common ground in the bank’s ongoing commitment to development operations (DevOps) philosophy – an application-development methodology that helps Agile teams work together more collaboratively and iteratively.
ANZ has previously reported “fantastic” results helping thousands of staff embrace DevOps, and Gray says this ongoing transformation helped her reframe the conversation around data in less territorial terms.
Rather than being something that had to be sacrificed and dumped into a generic data lake, data became seen as a common business asset to be strategically shared and guided.
Among other things, this involved rebalancing the bank’s skills development strategies and the makeup of its teams.
“In the Agile world, when you don’t have a lot of expertise you don’t know what you are hiring for,” Gray said, “and we would get a data analyst instead of what we really needed – a data engineer.”
“If you have lots of the first and none of the second, you end up with a dog’s breakfast.”
Business leaders also need a strong balance of skills, she added, noting that even the strongest data analytics specialists will struggle if they’re not being purposefully directed.
“You can do a lot of analytics training,” she says, “but unless the business people try to understand what data can do, they’re not going to understand the questions which create the work to be done.”
Driving cultural change with DataOps
These issues similarly dogged Lahti as she steered Hitachi Vantara through dramatic change that has drawn on DevOps’s flexibility and engagement to redefine the company’s data culture.
Her approach – which, like cybersecurity’s DevSecOps is a spin-off of the core DevOps philosophy – drew on the emerging field of DataOps to realign business staff around a simplified ‘edge core cloud’ model of the business.
By using virtualisation to find and source data where it is stored, rather than trying to extract it and dump it into a data lake, the goal was to “stop talking about the data and just do something with the data,” she said.
A six-week Data Value Envisioning (DVE) workshop united dozens of key business leaders to identify commonalities and direct the most strategic use of data.
Lahti catalysed the change by partnering with the company’s chief marketing officer – a great choice because “marketing is very data focused to begin with” – to deliver some early wins.
“He and I were each other’s friendlies,” she said, “and we pulled all the people we thought we would need into the room.”
“They were all willing to admit to the sin of hoarding their data repositories on their desktops and laptops.”
Brainstorming revealed 12 key hypotheses that promised real wins from the better use of data, and the team undertook a “scientific” evaluation that saw it evaluate them “one hypothesis at a time”.
By drawing on data scientists and complementary expertise from HR, legal, finance and other agencies, the CIO-CMO partnership identified $US28m ($A41m) in potential new revenue over 24 months.
The high-profile effort quickly gained support from business leaders who finally saw data as something that would serve them – and not something to which they would be beholden.
“We knew if we could get this right, we could repeat and reuse it,” Lahti said, flagging reusability of successful data methodologies.
“The cost of doing additional work with the data gets cheaper and cheaper to the point of almost being free.”
Embracing DataOps has brought the company a long way from where it started, to the point where data analytics is guided not by whimsy and gut feeling, but rather by drawing on reusable models of business engagement.
“You’ve created a safe place to have a conversation about the power of data,” she said, “and it doesn’t involve new infrastructure or new tools.”
“It involves everyone rolling up their sleeves, and getting into the data with subject matter groups and subject matter engineers; protectionism goes out the window, and it becomes a very collaborative conversation about data.”
