Kicking off the recent 11th annual MIT Chief Data Officer and Information Quality Symposium at the MIT campus, data management leaders convened for an inaugural panel discussion titled “Machine Learning and Advanced Analytics: What a CDO Needs to Know.” And what a chief data officer needs to know, it turns out, is a story.
While the idea of machine learning and related concepts like artificial intelligence (AI) got lip service, the conversation focused primarily on the need for data teams to find the story behind the data, in order to inform the analytics and then convincingly and compellingly tell that story in a way that demonstrates the issues and persuades the powers-that-be to take appropriate actions.
The first step, said the panelists, is to infuse the organization with a concrete sense of the value of data and its analysis.
“I’m a big believer that there’s no point in doing [data analytics initiatives] if you can’t provide economic value,” said Sanofi CDO Milind Kamkolkar, in his opening presentation: “What’s a CDO to Do in the AI World?” “It doesn’t matter what the price is … but if we don’t have the mindset to start valuing information and data, [then] as a product, we will fail.”
Kamkolkar went on to lament the state of resource allocation and risk tolerance related to his and other organizations’ data analytics initiatives. He noted that “Ninety percent of our resource allocation goes into low-value, modern, [everyday] work,” while only about 2 percent involves the hush-hush “urban myth” projects (mega-powerful analytics initiatives).
The solution Kamkolkar proposed involves broadening the backgrounds of the data teams by relying less on traditional data science backgrounds and more on storytelling backgrounds. Moreover, the people best suited for this job, the panelists agreed, are journalists—people who are already trained and experienced in concisely and convincingly telling a story to tailored audiences.
Kamkolkar noted that key aspects of data analytics—such as consumption data and the actual utility and actionability of data insights—are best served by what he calls “data journalism.”
“Unless you have journalists on your team, don’t try telling a data scientist to tell a story [about] what that data actually means,” he said. “Just hire a journalist [and you will] be amazed at the difference in the way you share information.”
Intersection Between “Quants” and “Translators”
“I love the word journalist. We need storytellers,” said co-panelist Nikhil Aggarwal, FinTech entrepreneur in residence at the iValley Innovation Center, in his subsequent presentation. “[This is] the intersection between the ‘quant’ and the ‘translator.'”
In data analytics in the enterprise, a “quant” is a pure-play data scientist who understands the numbers and the analysis, but may not have a firm grasp on how to translate those data and analyses into business solutions. A “translator” is someone who may have only rudimentary data analytics skills, but keenly understands and can speak to the business impact.
Kamkolkar and Aggarwal agreed that bridging the gaps between these two traditionally siloed roles with storytelling is essential.
“The intersection between marketing and analytics [presents] some core themes … that different parts of the organization are able to see,” said Aggarwal. “So how can we overcome these silos?”
Aggarwal added that the complexities of business problems are as much barriers as the complexities of data and analytics—particularly in his line of work: financial compliance. Consequently, in the interests of increasing agile decision making and collaboration, and busting silos, data journalists may need to translate for both the quants and the translators.
“The audience that you’re dealing with [is] used to 30-page written summaries of a compliance investigation,” said Aggarwal. “We’ve got to get the key focal points across to them.”
“Historically, data analytics groups have been closed groups … crunching numbers, doing their magic. The problem is that you can’t scale that,” Kamkolkar continued. “This entire capability needs to be disseminated and given freely.”
He said that another key benefit of democratizing analytics initiatives by including data journalists lies in the skeptical curiosity of these individuals, who will ensure that interesting and difficult questions will always be asked.
“I was in a meeting this morning where we were talking about bold initiatives we can take, but when you look at the bold initiatives, they’re not that bold,” Kamkolkar said. “I said, ‘Look, what if we got rid of our entire data management organization and just let it run completely on analytics. Would that be so bad?'”
The collective reaction, he related, was one of “dismissive perturbation. ” Nonetheless, Kamkolkar was steadfast in his philosophy.
“The reality is that we should be asking these questions,” he insisted. “Don’t be afraid to ask why you’ve got a certain initiative.”