Many organizations are struggling to get business value from their analytics. According to Gartner, through 2022, only 20 percent of analytic insights will deliver business outcomes. When it comes to AI, Gartner says 80 percent of projects this year will remain “alchemy, run by wizards whose talents will not scale in the organization.”
Setting up analytics projects or an analytics organization is one thing but deriving value from analytics is another. And with the COVID-19 pandemic disrupting economies around the globe, companies will likely take a close look at ROI when it comes to analytics and data science groups.
“It’s not about analytics. It’s not even about insights. It’s about impact. If you’re not making an impact, you’re wasting your time,” says Mike Onders, chief data officer, divisional CIO, and head of enterprise architecture at Cleveland, Ohio-based KeyBank.
Here, ruthless focus on business outcomes is key, as is the ability to rapidly prove analytics can have a business impact and then delivering results at scale.
“We actually work backwards from specific business outcomes we’re looking to achieve,” says Shri Santhanam, executive vice president and general manager of global analytics and AI at Experian. “Ultimately, ML [machine learning] and AI tend to be vehicles to get us to the ultimate goal, but really, what we talk about, what we share, what we drive with our customers is a better set of outcomes.”
Hurdles to analytics value
Institutional roadblocks are among the biggest obstacles to analytics ROI, says Brian Hopkins, vice president and principal analyst serving CIOs at Forrester Research.
“The problem we see is that CIOs’ data strategy involves a lot of business-level change, business process change, new organizational structures outside of IT to make decisions about definitions of data and make decisions about priorities, enforce data privacy policies, and other things that the CIO can’t control but have a cost.”
Hopkins says many companies are turning to their CIOs to offer a technology solution for deriving value from their data but aren’t seeing the bigger picture. Regardless of how much they budget for technology, they may need to invest twice as much in broad business change.
“This big chunk of firms who have this lofty vision of data strategy that’s fallen on the CIO in terms of budget to invest in the technology solutions are going to recognize that the cost of data strategy is much bigger than just IT and it involves a lot of business change,” Hopkins says. “It includes process change, application change, organizational change management, incentive changes.”
To get started, CIOs must create partnerships with business stakeholders. Ultimately, though, senior leadership must create incentives to drive those partnerships.
“You’ve got to incentivize business managers to care about data and to care about how well their data can be consumed by other lines of business,” Hopkins says. “That’s not something that most CIOs can readily wave their wand and change.”
Here, IT leaders shed light on how their organizations made the shift to analytics-driven impact and offer advice to those seeking to transform their data practices into business assets.
Onders says KeyBank has addressed this issue by making line-of-business stakeholders partner closely with the analytics team — and be accountable for the business outcomes of the analytics projects they request. For each project, KeyBank creates a one-page charter that describes the outcome the business seeks and the metrics for assessing that outcome. The charter lists the business sponsor, product lead, analytics lead, risk lead, and tech lead.
“The one-page charter says, ‘What metrics are you looking at?’ They have to list the metrics. If these analytics are going to change something, what metric is going to change? We’re going to hold you accountable that you’re expected to be this, by this date. How are the analytics going to get you there? It’s just a much more aggressive kind of metric-driven, impact-driven charter with a senior business leader that’s accountable for it,” Onders says.
Every two months the business leader must deliver a report showing how the analytics have affected the business outcome to justify continued investment in the project.
Experian’s data team also relies on close partnerships with the business to drive analytics initiatives, including an emphasis on getting “clarity on the business outcomes in a quantifiable way,” Santhanam says. But Experian’s analytics philosophy also allows for a more flexible, iterative approach.
“We start by doing things which don’t scale and hack through the business problem itself. This allows us to deviate from what might be a set of constraints, allowing us to operate very quickly and to experiment and understand where the leverage is,” Santhanam says.
Upskilling for analytics success
But for a business-IT collaboration on analytics to be successful, culture change is essential, as business pros must not only be versed in data-driven processes and technologies but also help fill gaps in areas where demand for skills outpaces supply.
Jabil is one company educating business pros and executives on what it takes to transform analytics into impactful business initiatives. The manufacturing services company has focused on becoming more data-driven for years. When it struggled to find data scientists, it created a Citizen Data Science program to help it mine the data at its disposal.
Jabil CIO Gary Cantrell says the two key components of Jabil’s push to become more data-driven have been a ruthless focus on addressing business problems and a closely related push for executive sponsorship.
Getting business leaders and senior executives on board was challenging but ultimately one of the most important factors in Jabil’s analytics success. As part of its Citizen Data Science program, Jabil created an executive-level training cohort that put key executives through a focused, two-day data science training program. The program helped executives understand the importance of becoming a data-driven organization and garnered their enthusiastic sponsorship as they started hunting for business problems to address with data. The program continues to reinforce executive sponsorship by inviting senior executives to participate in debriefings when each cohort finishes the program.
“We really started getting buy-in and traction across the executive team when they started seeing their organization’s problems being addressed and results coming out that could help them get better,” Cantrell says. “In all fairness, it took us a lot of selling on the front end. And then it took the better part of three years to get the senior executives excited about it. But now, the question the last two years has been, ‘OK, what are you doing with analytics? What’s next?’ It took a little while, but we finally got the message across by relating it back to the business, where they see the value.”
Breaking down data silos
Legacy data practices can also stifle an organization’s ability to transform its data into business value. The chief culprit? Data silos.
For the past several years, Bayer Crop Science has sought to apply machine learning and artificial intelligence to every aspect of its business. Precision agriculture has been a key focus. Michelle Lacy, data strategy lead for R&D in the Plant Biotechnology Division at Bayer Crop Science, says the company’s adherence to FAIR data — a set of guiding principles for scientific data management and stewardship published in Scientific Data — has been fundamental to its data-driven transformation, helping it break down silos of data.
FAIR (findable, accessible, interoperable, and reusable) is a sort of data bill of rights that says users should be able to find their data easily, users should be able to access the data they need when making decisions (while still following cyber security policies), data should be interoperable, and data should be reusable.
“It’s extremely important,” Lacy says. “It’s the foundation of our data strategy.”
Often, data developed by one group can be helpful to work undertaken by other groups. To make efficient use of the data, the various groups need to know the data exists and how to find it, and the data must be compatible.
“If you’re running various assays on a single plant, whether it’s field assays or you’re running different experiments in a lab, you’ve got to be able to bring that data together,” Lacy says. “You can think of it as a jigsaw puzzle and all these different assays you run are pieces of that puzzle. The project lead has to put those pieces back together.”
As with many transformations, the shift to a data-driven organization as hinges on trust: in teams members, in new processes, and that insights derived from data will have a positive impact on the business.
At Experian, “four pillars” guide every machine learning and AI project, Santhanam says: performance, scaling, adoption, and trust.
Historically, Santhanam says, the issue of trust has limited what banks would do with their analytics models.
“A lot of the positioning models have been fairly simple logistic regression models for the very reason that the confidence in building something more complex and opaque creates a level of risk which feels outside the risk appetite of organizations that are heavily regulated. What we are seeing, though, is both the regulatory framework and businesses in this space are recognizing the value of more complex algorithms and more complex techniques and taking more of a staged approach of responsibly moving into that space with explainable AI frameworks,” Santhanam says.
“Ultimately, driving impact requires all four of those things, and it’s quite easy to lose your way if you don’t have the business outcomes in mind as your goal,” he adds.