IT Computer

SAS’s Nevala drills down into what it takes to achieve analytic success

It is a connundrum executive teams of numerous organizations who have hit major road bumps in their analytics development journey must surely discuss among themselves or with others: Why do some rollouts fail miserably while others succeed?

The answer to the question, said Kimberly Nevala, strategic advisor and advisory business solution manager with SAS, can be crystalized in six key attributes that companies who make “good use of analytics” adopt and practice.

SAS executive Kimberly Nevala,  delivering the keynote speech yesterday at the Analytics Unleashed live and virtual event.

In a keynote speech yesterday at the second annual Analytics Unleashed event, organized by IT World Canada and sponsored by SAS, Informatica and shinydocs, Nevala detailed six attributes that organizations need to have to not only achieve success, but to adapt to changing times.

Attribute One: Those firms that succeed in using analytics and artificial intelligence (AI), she said, focus on solving a broad spectrum of problems, full stop, end of story. “They are applying analytics and AI to problems that are both big and small. And in fact, the companies that are most mature report that the balance between use cases that you might consider operational, and those that are more strategic – things that are focused on operational efficiencies, versus creating new products or services – is about 50-50.”

The takeaway, she said, is, “companies who do this well no longer think about and plan for their data and analytics strategy to be separate from their business strategy.”

Attribute Two: Successful companies already use a broad spectrum of tools and as a result, are the least inclined to be distracted by the new bright and shiny objects: “They use the most simple, well proven techniques they can to solve any problem. And they do not spend a lot of time going back and re-architecting or redesigning something that already works, just because there’s a new method that could also work,” said Nevala.

“We might not take our old approach to forecasting and replace it with a machine learning model unless I can show a germane business impact and reason for doing it now. Why do I mention that? It’s important because they are not spending a lot of time just retreading existing ground.

“Now they have the headspace to go out and find new analytic problems to solve because they are not trying to make incremental, non-germane improvements in areas that are already doing well.”

Attribute Three: The successful organizations invest incrementally and mindfully in infrastructure, she said. What that means is that the “analytics and data infrastructure strategy is closely tied to their operational and transactional infrastructure strategy. And what this looks like is that companies that, for instance, are early adopters to the cloud, are not running to lift and shift every analytic workflow and all the accompanying data immediately to the cloud.

“They’re being mindful about the analytic workloads that make sense, and would benefit from the capabilities that are available in the cloud. It means that they invest in developing a robust blueprint for modern data pipelines, but they don’t try to move every data stream onto it before people start using it. They prioritize those data streams based on the use cases and actual usage and value in the organization.”

Attribute Four: They are big believers in mandatory AI and analytics training for every staff member. Nevala referenced an Accenture study entitled The Art of AI Maturity: Advancing from Practice to Performance that revealed that only 12 per cent of companies can be described as AI achievers. “On average, these companies are saying they can relate 30 per cent of revenue gains to their AI projects overall. That’s a staggering amount, but what I found really interesting was that 78 per cent of those AI achievers have mandatory training for employees at all levels of their companies.”

Training, she said, is not about teaching people number sense and understanding statistics, but teaching them about “analytic recognition so that people in your organization can actually know and identify the types of questions and the problems they can answer and the problems they can solve with analytics.

“Why is this important? It’s important, because it increases the surface area, if you will, the number of people who can identify problems we can apply analytics to. And because these people are identifying problems they care about, it increases the likelihood that the solution will be adopted.”

Nevala also stressed that simply having the tools in place will not guarantee success. As proof of that, she recalled a quote from the Scottish poet, novelist and literary critic Andrew Lang, who famously once said ‘politicians use statistics like a drunk uses a lamppost – for support rather than illumination.’

“It sounds like a joke; however, there was a recent study and in it, only 22 per cent of the decision makers surveyed said they use the insights and data that are provided to them when they are making decisions.”

Attribute Five: Successful organizations implement a strategy that involves decision intelligence (DI), a discipline that factors in data output from machine learning (ML) and AI advances. “Like so many other things, we have to develop the muscle and the skill in our organization to make good decisions about using information,” said Nevala.

“Frankly, I could probably use this in my day-to-day life as well. But what this means is that we are going to be very deliberate about identifying the decisions that we want to inform or make with analytics. And we are also going to define how we will make the decisions using the information that is provided.

“And then we’re going to monitor the outcomes of those decisions. To be clear, the point of DI is not to eliminate human judgment, the point of it is so that we are clear about how we apply the machine prediction. How will the human use that machine prediction when they are making a decision?”

Attribute Six: The final attribute revolves around a single word – governance. “The standard approach to governance, or thinking about governance, is that it is going to stymie innovation,” she said. “I would argue exactly the opposite, that if done well, particularly now, when we have to be attentive not just to risks, but increasingly to rights, it is the key to unlocking innovation.

“If we do governance right, (it) is about enabling critical thinking, and allowing people to make decisions in the face of uncertainty.”

In the end, said Nevala, analytic tools and platforms should be considered as a means to an end: “Now there is no question that low-code, no-code, and data scientists are very, very important. And they can get a lot more people in your organization developing insights, models, etc.

“But you should be under no illusion that the majority of your employees want to roll their own analytics. They don’t. And they won’t, and nor does their job or their function likely require them to, moving forward. But this doesn’t mean that they’re not interested in doing better with insights and outcomes that a model can give them.”

She observed that, like kids whose parent hides the spinach in their kids’ cheesy lasagna, “they prefer that those insights are delivered to them in the context and in line of their existing business process flows and workflows, not as a separate tool. Organizations that assume that analytics and AI are going to be self-serve for everybody may find that analytics and AI are self-serve and used by nobody.”