BACKGROUND IMAGE: stock.adobe.com
Augmented analytics vendors are starting to infuse AI capabilities into various types of workflows. These promise to streamline analytics processes, generate more useful results and enable more users to conduct their own analytics.
Gartner coined the term "augmented analytics" to describe this trend, which involves weaving AI into analytics, data preparation, and data science and machine learning workflows, and integrating these different capabilities together. All the leading analytics vendors, including IBM, Tibco, Microsoft, Tableau and Qlik, are adding this kind of functionality into their platforms.
"Most of the BI vendors are now injecting AI into their product, so it is pretty much becoming the cost of doing business," said Wayne Eckerson, principal of Eckerson Group, a research and consulting company focused on business intelligence and analytics.
Applications of AI in BI and analytics tools include improving self-service features, automating analysis, surfacing recommendation for relevant analytics and enabling natural language queries and responses. Vendors are also adding integration with data science programming languages like R and Python and then simplifying the workflow of running those models inside the analytics toolset.
But the traditional vendors also face challenges in extending their offerings beyond the desktop in order to take advantage of new augmented analytics workflows, Eckerson said. Applications like Tableau, Power BI and Cognos were designed to run on desktops, but many augmented analytics use cases involve moving the heavy processing to more powerful and connected servers running in a cloud.
"It is painful for them to rearchitect around a server-based environment with a thin client," Eckerson said.
He sees a possibility for a shakeup in the industry, akin to the way that tools like Tableau, Qlik and Spotfire initiated a previous revamp with more iterative, visual analytics. This time around the charge is being led by startups like Yellowfin and ThoughtSpot. These new tools bypass the desktop phase of analytics, which Eckerson said makes them better situated for the future.
The newer augmented analytics vendors support natural language queries, automated insights, data management capabilities and data science integration in a more coherent manner than legacy vendors. "It is a lot easier when you don't have a lot of the legacy baggage to support," Eckerson said.
In contrast, established vendors have launched different products for various aspects of augmented analytics workflows. For example, IBM provides Cognos for analytics and IBM Watson Studio for data science. Similarly, Tibco provides Spotfire for analytics, Jaspersoft for reporting and Tibco Data Science for data science and machine learning.
Augmenting the integrated platforms
The big analytics platforms, including Microsoft Power BI, IBM Cognos, Tibco Spotfire, Tableau and Qlik Sense, provide a one-stop shop for the most popular types of analytics today including reporting, dashboards, visual exploration, ad hoc reporting and predictive analytics. They also bring to bear a common set of architectural services from management, administration, security and governance, so you can deliver a complete environment using one toolset from a single vendor.
All these augmented analytics vendors are also starting to add data preparation, data science and machine learning features. In theory, at least, this should mean that enterprises will bring in augmented analytics on top of these platforms in a way that supports a common look, feel and interaction framework. But in practice, most companies already have a heterogenous environment.
"As much as they want to standardize on one analytics platform, they cannot," Eckerson said.
He recommends that enterprises focus on platforms that are open and extensible so that other analytics tools can query their reports and semantic layers. It is also important that these platforms have published and documented APIs, because that makes it easy to access functionality from within other applications the enterprise plans to use.
Eckerson also sees vendors adding better data management capabilities in order to improve performance or persist data. This includes better tooling for newer data sources like JSON formats, which are widely used for web apps. He recommends enterprises look for a set of connectors that make it easy to pull or extract data from JSON sources, or query from them directly, since more business data is being created in JSON formats as nested data sets. Few tools have this capability today, but Eckerson believes that this will be an important differentiator for augmented analytics platforms in the near future.
Improving analytics processes
Dave Menninger, senior vice president and research director at Ventana Research, said business analytics professionals should evaluate two general categories of augmented analytics capabilities: 1) doing new types of analyses and 2) assisting with the existing use of BI products.
In the first category, one of the most sought-after capabilities is natural language explanations of analyses, such as why sales have increased or decreased. This also includes automatically identifying the key drivers in different analyses. Leading analytics vendors are starting to add some natural language querying and natural language generation techniques to their tools, but these are still immature.
"The natural language explanations are not yet as 'natural' as they could be," Menninger said. "However, they do a fairly good job at identifying key drivers."
In the second category, vendors are applying AI and machine learning to help identify how to work with data. This includes things like deciding which tables should be joined and how. It also includes recommending new analysis techniques based on your past usage patterns and your similarity to other users.
Menninger believes the major weaknesses of the current augmented capabilities are overlooked. Augmented analytics vendors are currently focused on reactive historical analyses rather than improving prescriptive forward-looking analyses.
"Sure, there are some forecasting capabilities, but no one is offering serious churn analyses or predictive maintenance via augmented analytics," Menninger said
He cautions enterprises to set their expectations appropriately. "If you expect that you will be able to replace or avoid hiring a data science team, you will be greatly disappointed," he said.