Narrowing the AI-BI Gap with Exploratory Analysis
The worlds of AI and BI occupy distinct places in the analytics continuum, which is most often understood with concepts like descriptive analytics, predictive analytics, and prescriptive analytics. Users can leverage descriptive analytics and BI tools to explore what happened in the past, while predictive analytics makes use of ML models trained on real-world data to generate an educated guess about what will happen next.
However, the lines separating these two camps are getting more blurry by the month. For years, Gartner has talked about how BI tool vendors are adding more ML and AI capabilities to their wares. In its latest Magic Quadrant for Analytics and BI Platforms, the firm talked about how the next generation of “augmented analytic” products will bring ML and AI to bear on things like data prep, query generation, and insight generation. At the same time, we’re seeing vendors that develop data science development and AutoML environments embrace SQL as a necessary ingredient for data preparation, manipulation, and interrogation.
One of the newer vendors in this segment is Tellius, an analytics startup based in Reston, Virgina. According to Tellius CEO and founder Ajay Khanna, there’s a functionality gap between what the top BI and AI tools can do.
“On the BI side, we see tools like Tableau, PowerBI, and Qlik do a pretty good job of visualization, but that’s kind of where they stop,” Khanna says. “And most of the business users and analyst we talk to want to go beyond visualization to understand why things are happening, but that process is very manual. That’s looking from the BI lens.”
When Khanna views the situation through the ML and AI lens, he finds technical complexity that exceeds the abilities of your regular business user or analyst.
“If you look at DataRobot–amazing company,” he continues. “But look at the interface. It still has a technical names like ‘area under curve’ and ‘confusion matrix.’ Those are not things which the typical business user or the analyst understands. So we see this huge insights gap created by these two silos. Our vision is to bridge that gap.”
Khanna and his colleagues built the Tellius platform to augment the analytics capabilities of mainstream BI tools by bringing AI and ML technologies and techniques to bear on big data. The platform, which was built using Apache Spark and features an in-memory columnar database, was designed to enable regular users to explore massive data sets without manual, SQL-based slicing and dicing.
For starters, it leverages natural language processing (NLP) capabilities to present a simpler search-style interface for users. The software converts typed questions into targeted SQL queries that execute against data stored in its columnar database and returns the results in a matter of seconds.
“Ok I want to look at revenue by state. The system knows you asked for ‘by state’ and it will make a chart,” Khanna tells Datanami in an interview earlier this year. “If you say ‘Give me revenue by channel and product,’ it will give you those answers. That’s the beauty of this. Our customers are doing this on billions of records and still getting second-level response, which is unbelievable but absolutely possible. And you don’t have to limit yourself with the questions.”
Thanks to its distributed architecture, Tellius is able to query larger data sets than can fit in a Tableau or Qlik environment, Khanna says. That eliminates the need to pare the data down to fit it into a BI tool for exploratory analysis, he says.
“Spark is a very distributed technology, which means if you’ve got terabytes of data, we can load it in our system in a distributed manner and still give you blazing fast results,” Khanna says.
The system can also leverage its natural language query (NLQ) capability to generate questions all on its own. If a user is just starting out, she can ask the system what she can ask, and Tellius will offer suggestions. In addition to asking questions in a NLQ manner, it generates answers using natural language generation (NLG).
Tellius also supports the next level beyond simply searching the data, which is asking more targeted questions, like why something happened. That’s not a simple thing to automate, but Tellius gives users tools to help narrow down the huge number of possibilities into something more manageable.
“Okay, I want to know why this changed, why this dropped by 66% between April and March,” Khanna says. “The system runs this analysis and goes through, slices and dices millions of combinations, which can be thousands of lines of SQL code.”
Finally, Tellius also sports basic AutoML functionality, including the capability to generate a model based on data stored in the database. It’s not quite to the level of what a full-blown data scientist would like, or what tools like DataRobot can offer. But it makes the citizen data scientist more useful to his or her oganization, Khanna says.
Tellius, which was founded in 2015 and has raised $17 million to date, has been selling its software for a couple of years, and so far, it’s gained the most traction in the consumer processed goods (CPG) and pharmaceutical industries, which are places that typically do not have an abundance of data science talent available, Khanna says.
“They have a lot of data, but they don’t have data scientists,” he says. “They still have to do all the slicing and dicing, and that’s where we see the biggest value-add.”
Last week, Tellius launched Live Insights, which is a new capability designed to piggyback on the processing power of customers’ existing cloud data warehouses. The company says Live Insights “accelerates the discovery of these business insights from terabytes of unaggregated data by analyzing the data inside cloud data platforms,” such as Snowflake, Amazon Redshift, and Google Cloud BigQuery.”
Live Insights is currently available to select customers and will be generally available in Q1 2022, the company says. For more information, see www.tellius.com.