Augmented analytics is a term coined by Gartner to describe the integration of natural language processing, natural...
language generation, text mining and automated data processing capabilities into BI systems. All the major BI vendors are buying or building these capabilities into their BI platforms in order to make it easier for enterprise customers to democratize analysis.
"I believe the top players in BI understand that augmented analytics holds both a promise and a threat for them," said Micha Breakstone, co-founder and head of R&D at Chorus.ai, a conversational analytics service. The promise is that augmented analytics tools will make BI more accessible to nontechnical business experts. The challenge top BI vendors face is that their BI platforms may be replaced by these automatic systems that proactively surface actionable insights more efficiently.
As a result, top BI vendors are focused on maximizing the potential upside by simplifying interfaces and reducing friction of highly technical interfaces and proprietary languages, Breakstone said. At the same time, they are working to minimize risk by both providing deeper analysis capabilities for data scientists and designing their tools to be more insight-driven.
Mark Palmer, general manager of analytics at Tibco Software, said, "Business analysts today want the benefits of AI without having to become a Ph.D. So BI vendors are building tools that make it easy to bring machine learning to your fingertips wrapped by an easy-to-use interface that end users can use to explore." This can include everything from easier interfaces for building machine learning models to adaptive AI, where the results can be adjusted based on changing conditions. For example, when monitoring a stream of transactions and streaming weather feeds, a predictive model can suggest real-time adjustments to sales campaigns when it knows it's going to be cold, wet or sunny in a geographic region.
Aspects of augmented analytics tools
In Gartner's model, augmented analytics uses machine learning to automate data preparation, insight discovery and insight sharing for business users, operational workers and citizen data scientists. Startups and large vendors could disrupt leading BI and analytics, data science, data integration and embedded data analytics vendors.
Some of the key capabilities of augmented analytics tools include augmented data preparation, augmented data discovery and augmented data science and machine learning. Augmented data preparation focuses on automating the ingestion of data into analytics systems. It includes processes like data profiling, ensuring data quality, modeling data, adding metadata and storing it in catalogs.
Augmented data discovery involves helping users find relevant data. This includes automating, visualizing and narrating relevant findings. Machine learning helps reduce the skills required to build models or write algorithms.
Augmented data science reduces the skills required for business experts and data scientists to test out new hypotheses. It includes processes like automated machine learning-model feature selection that streamlines processes around generating, deploying and managing advanced analytics models.
Machine learning is key
The key trend is that top BI vendors are exploring different ways to weave machine learning into different aspects of BI. Some vendors believe this allows enterprises to shift their focus from discovering trends after the fact to providing automated and hidden insights to business users at the right moment.
Stephen BlumCTO at PubNub
Nic Smith, global vice president of product marketing for cloud analytics at SAP, said, "The landscape of BI is changing from passive to active, in which insights are automatically generated based on data vs. user-created reports and dashboards. This trend is largely driven by the advancement of machine learning and cloud technology, as well as the increased speed of business."
One key driver is that enterprises are looking for innovative ways to make sense of growing streams of data from e-commerce, mobile apps, social media, IoT and call center analytics. "Business intelligence vendors are realizing that only providing functionality for teams to manually query data isn't good enough," said Stephen Blum, founder and CTO of PubNub, a data management API provider. Augmented analytics provides a new way to analyze, process, crunch and derive insights from this data.
What to look for in augmented analytics tools
John Hagerty, vice president of product management for business analytics at Oracle, suggested that analytics managers consider support for the following five augmented analytics capabilities in their conversations with top BI vendors:
- Recommendation: The system should recommend how data could be mashed up, what the best visual for specific data is, how to enrich data for deeper analysis and understanding, and how to clean and prepare data for broader use.
- Insight generation: Augmented analytics tools should be able to drill deeper into what drives specific performance. Analysts may bring their own ideas (biases) to the process and will seek out results to support their hypotheses. Insight generation uses algorithms to describe the data, identify key drivers, explain what segments influence the outcomes and which situations are outliers, behaving differently than the anticipated results. This needs to happen seamlessly, with results shown to the analyst to uncover potential hidden insights and drivers.
- Natural language processing: Users should be able to use the power of language -- text-based and/or voice-enabled -- to interact with the data in a conversational mode.
- Natural language generation: BI tools with augmented analytics capabilities should be able to use the power of language to narrate performance results in an interactive way. As a user digs deeper, the narration adjusts to explain the context of what they are looking at.
- Prediction: Augmented analytics tools should be able to easily forecast a trend, identify outliers and cluster groups of like values with the click of a button. This also includes using algorithms to train predictive models for all manner of use cases, such as churn, attrition or behavior. It should also be able to incorporate those predictions into the body of analysis.