Tibco Eyes ‘Data Science for Ops’ with Spotfire Upgrades
Data management and business intelligence vendors continue to push deeper into the cloud with a growing list of open-source tools while adding automation features such as machine learning and natural language queries designed to provide what has come to be known as augmented analytics.
That cloud-native, open-source, AI-driven approach sums up the strategy of Tibco Software, which released the latest version of its flagship Spotfire data visualization and analytics software last November. It rolled out Spotfire enhancements last week during a company event in Chicago, including native support for interactive analytics on large data sets running on frameworks like Apache Spark and the Google BigQuery data warehouse.
The upgraded platform adds an “AI engine” that provides explanations as well as recommendations, making AI “part of the user experience,” Michael O’Connell, Tibco’s chief analytics officer, said in an interview.
That’s consistent with Tibco’s overall strategy of embedding AI into analytics applications via AutoML and other methodologies. The goal is making these smarter analytics tools easier to use for business analysts as well as data scientists.
O’Connell said the Spotfire upgrades represent an “AI on demand” service for analytics users, especially for edge and other event-driven applications. The company announced a new open source project last week allowing developers to work with streaming data. The “streams designer” falls under its Project Flogo, an open source repository for event-driven apps.
Tibco, which also last week announced a cloud partnership with Microsoft Azure, is offering starter kits in the cloud, including a web client that adds support from its data visualization tool that allows users to refresh data and view disparate data sources.
Meanwhile, Tibco Labs is pushing “foundational AI” across the Bay Area company’s product line.
The Azure partnership reflects efforts by analytics vendors to take advantage of the enterprise shift to multi-cloud deployments. Tibco earlier partnered with Amazon Web Services (NASDAQ: AMZN)
for cloud machine learning applications as well as micro-services and event-driven APIs.
“We recognize that many customers already have a strong footprint with Azure, that developers need less complexity when building solutions and that project managers and collaborative teams need their tools to work efficiently,” Matt Quinn, Tibco’s chief operating officer, said during last week’s Chicago shindig.
The cloud partnership also makes sense since both Microsoft (NASDAQ: MSFT) and Tibco are among the leaders in the emerging category of augmented analytics. According to market watcher Gartner Inc. (NYSE: IT), Microsoft leads a crowded field in the segment that combines machine learning and other automation with analytics and business intelligence tools. Tibco was listed among the industry “visionaries” in rankings released in February.
O’Connell, who described his “offensive position” within Tibco as equal parts customer-facing and product development, said Spotfire continues to gain traction in industrial and manufacturing applications where users can leverage “data science on operations.” Those machine learning efforts build on the Flogo initiative and coordination with Tibco Labs as the software vendor targets edge applications.
Those enhancements would improve Spotfire’s ability to handle event streams and time-series data that could be used in manufacturing settings for applications like equipment anomaly detection. Among the requirements, O’Connell explained, is dealing with what he called “wide” time-series data generated by industrial edge devices. In one example, a manufacturer was generating 6 million columns of data. Six million “columns, not rows,” he added. “It was off the charts.”
Hence, Spotfire is being tuned to sort through hundreds of data points a second. For applications like anomaly detection, “You don’t need all those data points [to retain] the signal in the data,” O’Connell said.
In an operational setting, an anomaly in sensor data would generate an alert, prompting a manufacturer to assign a technician to fix a predicted problem. At the same time, the model used to identify a problem could be retrained to incorporate how an anomaly was spotted. That, in turn, addresses the tougher problem of anomaly detection in a multi-dimensional space.
The point is to provide an analytics platform that allows operators to spot a problem faster and fix it. “That’s the value proposition,” O’Connell said.