Organizations are increasingly recognizing the value of data science and advanced analytics. However, while investments in these technologies are growing, many companies still struggle to achieve meaningful business impact from their data initiatives, regardless of their size.
I recently had the privilege to interview Howard Dresner and Brian Lett with Dresner Advisory Services on the critical and growing role of advanced analytic techniques in enterprises. They shared surprising results from their research, best practices, and actionable steps for organizations wanting to maximize the results they get from their data.
The undeniable need for data science capabilities
When I asked Howard and Brian about enterprise interest in data science, Howard said 47% of organizations say that data science is either critical or very important. And 52% are actually using data science and machine learning in some fashion today. But traditional approaches don’t address the challenges of the huge and growing volume of data that gets “stored and ignored.”
“Data doesn’t improve with age,” Howard Dresner said — a fact that anyone in the data industry can attest to. “If this data was properly analyzed, it would add huge value to the organization.”
To maximize the value of their data, organizations must distill and make sense of their data — while it still matters — to better understand key relationships and influences. But to do that, they need advanced analytics techniques.
However, only 40% of users can access any data-driven insights. Because of this insights gap, “there’s a lot of opportunity for organizations to improve data-driven decision-making across the organization,” Howard said.
Must-have data science capabilities
Not all organizations are in the same place with their model and data science deployments. Some have dedicated data science departments; others use application or service providers. But even in smaller organizations, more than half view data science as critical or very important, according to Dresner’s research.
So, for an organization of any size, which capabilities are most important? Howard explained that organizations must seek out organic platforms, or in other words, platforms that “work together with all the various components and features [in the data science life cycle] with a complete set of tools and routines that allow users to experiment and easily iterate. Users want a fluid, cohesive, and consistent environment where they can easily build, test, and manage their analytical workflows. This includes access to any number of data sources, both cloud and on-prem.
“This also means supporting different kinds of users, basic to advanced, with a combination of low-code, no-code, and full-code alternatives — because users, over time, will graduate as they learn more about the platform and may expand into those different modes of operation.”
Brian added that organizations believe data prep capabilities are essential to the data science process. “If you don’t get this part right, it just becomes a bottleneck down the line.”
Wherever possible, Howard advised organizations to automate the results of these models as inputs into operational processes and workflows. “The only way to achieve the digital enterprise and leverage all this data is going to be through automation.”
How to start deploying better data science practices
To succeed, organizations “need to become hyper-decisive. They need to be able to very quickly leverage these huge amounts of diverse data and transform that into relevant and actionable insight. And this is really at the core of becoming a digital enterprise,” Howard said.
Brian added that it’s a progression. “Not every organization is going to be a type of runner right away. They’re going to need to crawl before they can walk before they can run. Platform vendors are going to provide capabilities that allow them [organizations] to evolve. And as we mentioned earlier, making these capabilities available throughout the organization and not just having them focused solely on the data science team is going to be important for not only having the capabilities there but also making it more of a cultural element where there’s more data-driven decision-making.”
Having approachable, easy-to-use analytics available for multiple users across the enterprise is so important. Equally important is collaboration between data science and business stakeholders. Brian explained, “Ideally, regardless of which function you’re talking about, it’s the data scientists working very closely with the key business stakeholders, whatever the functional area may be. Look to leverage the subject matter expertise that exists within the business functions. … Use and leverage that collaboration all the way through the whole lifecycle for defining, testing, deploying, monitoring, updating, managing, and even setting the goals for the modes.”
Brian also gave some advice for how organizations should think about building models. “We believe every DS and ML model needs to start with framing and defining the business problem or business decision that needs to be made. You don’t want to start by thinking about things like features and models. … Because it’s really hard to create business value from a solution that’s in search of a problem.”
Here are a few things to help accelerate your journey to insights