GRAPEVINE, Texas -- "It's such a consequential time to be a data and analytics leader," said Rita Sallam, research...
vice president at Gartner and master of ceremonies of the recent Gartner Data & Analytics Summit 2018 event. Consequential because companies deemed info-savvy are valued at nearly twice the market average, she said, citing Gartner research. And consequential because data, as regarded by the information experts attending the event, is under attack.
"Just as fake news became a viable political weapon -- and make no mistake, fake news is fake data, which makes it our problem -- ensuring data quality, providing a foundation of trust, just became job No. 1 for everyone in this room," Sallam said.
In a kickoff presentation notable for connecting the job of data and analytics leaders to national affairs -- and for espousing a political point of view -- Sallam and her Gartner colleagues warned attendees they must overcome four "tough challenges" as they strive to help their companies capitalize on data. To succeed, they will have to:
- Establish trust in the data;
- Promote diversity -- of people and skills, as well as types of data;
- Manage complexity through automation; and
- Develop data literacy programs.
Gartner's advice on how to meet these challenges, like Sallam's callout to fake news, contained some provocative nuggets. The segment on diversity, for example, made a case for hiring people with autism and cited the Supreme Court case involving a criminal sentenced by algorithm as a cautionary tale about bias in data science. The literacy segment questioned the sanity of allowing employees to use corporate data to make business decisions without first passing a data proficiency test.
Another eye-opener from the keynote? Bimodal IT, the term coined in 2014 by Gartner to signify the two-tiered IT approach best-suited for supporting digital businesses, is no longer sufficient.
"For the last three years, we've come on the stage and we've recommended a bimodal strategy -- Mode 1 for well-established production content and Mode 2 for agile innovative content," said Kurt Schlegel, research vice president at Gartner, adding that "most of us have executed on that strategy reasonably well." IT teams have maintained Mode 1 systems to run the business and added a Mode 2 layer of small teams empowered to innovate.
Kurt Schlegelresearch vice president, Gartner
"That sounds terrific, but we have hit a wall," Schlegel said. Integrating the two modes has proved difficult to do. He pointed to a new mode of working inspired by crowdsourcing projects like the XPrize and supported by automation tools that use artificial intelligence techniques such as machine learning and natural-language processing.
"That is how we must scale the value of data and analytics, by shifting from manual processes done by a few to automated processes done by the many," he said.
Here's a breakdown of the four challenges and a synopsis of how Gartner believes each should be addressed.
Data and analytics leaders should not underestimate the current risk of people intentionally disseminating misleading information, Schlegel said. "This is by no means contained to politics. In a world where scammers are using much more sophisticated phishing techniques, they can spoof almost any company or situation."
Companies tend to treat all data like it comes from a system of record. "Very little actually does," he said, which puts enterprises at serious risk. "We must be able to verify our data to better inform all the decisions we make."
The antidote to data blindness is metadata -- but not the level of metadata created by data custodians and compiled in business glossaries nobody reads. Data and analytics leaders need to crowdsource data veracity.
"We need mainstream users to tag what they do and, as a result, provide a record of data lineage along the way," Schlegel urged, noting that non-data specialists routinely tag their activities on social media. "We could do the same at work," resulting in a "more dynamic method" of trusting data.
"I'm here to communicate how establishing all kinds of diversity and inclusion as a core principle is not only the right thing to do; it is absolutely essential to scaling the value of your data and analytics programs -- to driving innovation and minimizing bias in algorithms," said Carlie Idoine, research director for business analytics and data science at Gartner.
A 2018 McKinsey report found that organizations in the top quartile for ethnic and culturally diverse executive teams were 33% more likely to have above-average profitability than those in the bottom quartile. Idoine said Gartner's most recent business intelligence (BI) survey also found that diverse teams derived a "higher achievement of business benefit" from their data analytics initiatives. "There's substantial research that shows diverse teams perform better than homogenous teams," she said.
Another reason to broaden hiring practices is the shortage of data analytics talent -- a crippling gap that Idoine said will require data and analytics leaders to look beyond gender and ethnic parity.
"Hiring practices, let's face it, are a bit biased toward the shiny, happy people of the world. Not everyone fits that mold," she said. "By sourcing teams with a diverse array of talent inclusive of gender, race, age and work styles, introverts and extroverts, and absolutely inclusive of the people on the autism spectrum, we can reach much higher performing data and analytics teams."
Diverse sources of data, coupled with diverse teams, will also mitigate building algorithms skewed by human bias, Idoine said, referencing data scientist Cathy O'Neil's work on destructive algorithms.
To deal with the diversity of data required by today's businesses, Schlegel said that data and analytics leaders will need to expand and speed up their analytics programs.
"The challenge is creating a data and analytics program that finds that ideal balance between agile creativity and enterprise scalability," he said. It won't be done through bimodal IT nor by using the sequential approach of traditional BI efforts, where tasks are completed and handed off to the next expert.
"That assembly-line approach, while efficient, is just not creative and collaborative enough for our needs," Schlegel said.
Instead, analytics organizations should be looking to solve complex problems by placing bets on many small teams empowered to build low-risk, low-cost prototypes. "This is a crowdsourcing method of scaling value, and it magnifies the odds of doing something great, something powerful enough to tame the complexity our organization faces," he said.
Develop data literacy
We don't allow people to drive metal boxes at high speeds past pedestrians without having a driver's license. "Isn't it crazy that we grant users access to mission-critical data without any kind of certification? As Uncle Ben told Peter Parker, 'With great power, comes great responsibility,'" Idoine said.
To scale their analytics programs, companies will need to recruit so-called citizen data scientists from employee ranks and support them with the automated tools designed to help nonprofessionals derive business insights. But it's imperative to add some certification for accessing data, Idoine said. "Just as we have different types of driver's licenses for different types of vehicles, we will have different types of education and training and testing for different types of users, consumers and citizen data scientists."
Data literacy is essential to thriving in a digital economy and it will require "committed professional development form the board room to the break room," Idoine said.
The monks of the Middle Ages "did great work," she said, in preserving data literacy by recording the information of the day and transcribing the classics, but it took a technological breakthrough -- the printing press -- to build literacy among the masses.
So, too, data literacy will require a technology catalyst, she said. Data discovery tools, once thought to be the answer for democratizing analytics, have proved too manual. Instead, she urged attendees to embrace what Gartner calls augmented analytics, described by Gartner as using data discovery tools augmented with machine learning and natural language to automate analytic insights. Automation will help turn users into responsible citizen data scientists -- assuming, that is, attendees are willing to help create a culture of data literacy.