No-code AI analytics may soon automate data science jobs

SparkBeyond, a company that helps analysts use AI to generate new answers to business problems without requiring any code, today has released its product SparkBeyond Discovery.

The company aims to automate the job of a data scientist. Typically, a data scientist looking to solve a problem may be able to generate and test 10 or more hypotheses a day. With SparkBeyond’s machine, millions of hypotheses can be generated per minute from the data it leverages from the open web and a client’s internal data, the company says. Additionally, SparkBeyond explains its findings in natural language, so a no-code analyst can easily understand it.

How companies can benefit from AI analytics data automation

The product is the culmination of work that started in 2013 when the company had the idea to build a machine to access the web and GitHub to find code and other building blocks to formulate new ideas for finding solutions to problems. To use SparkBeyond Discovery, all a client company needs to do is specify its domain and what exactly it wants to optimize.

SparkBeyond has offered a test version of the product, which it began developing two years ago. The company says its customers include McKinsey, Baker McKenzie, Hitachi, PepsiCo, Santander, Zabka, Swisscard, SEBx, Investa, Oxford, and ABInBev.

One of SparkBeyond’s client success stories involved a retailer that wanted to know where to open 5,000 new stores, with the goal of maximizing profit. As SparkBeyond CEO Sagie Davidovich explains, SparkBeyond took the point-of-sale data from the retailer’s existing stores to find which were most profitable. It correlated the profitability with data from a range of external sources, including weather information, maps, and geo-coordinates. Then SparkBeyond went on to test a range of hypotheses, including theories such as if three consecutive rainy days in proximity to competing stories correlated with profitability. In the end, proximity to laundromats correlated the most strongly to profitability, Davidovich explains. It turns out people have time to shop while they wait for their laundry, something that may seem obvious in retrospect, but not at all obvious at the outset.

The company says its auto-generation of predictive models for analysts puts it in a unique position in the marketplace of AI services. Most AI tools aim to help the data scientist with the modeling and testing process once the data scientist has already come up with a hypothesis to test.

Competitors in the data automation space

Several competitors, including Data Robot and H20, offer automated AI and ML modeling. But SparkBeyond’s VP and general manager, Ed Janvrin, says this area of auto-ML feels increasingly commoditized. SparkBeyond also offers an auto-ML module, he says.

There are also several competitors, including Dataiku and Alteryx, that help with no-code data preparation. But those companies are not offering pure, automated feature discovery, says Janvrin. SparkBeyond is working on its own data preparation features which will allow analysts to join most data types — such as time-series, text analysis, or geospatial data — easily without writing code.

Since 2013, SparkBeyond has quietly raised $60 million in total backing from investors, which it did not previously announce. Investors include Israeli venture firm Aleph, Lord David Alliance, and others.

“The demand for data skills has reached virtually every industry,” said Davidovich in a statement. “What was once considered a domain for expert data scientists at large enterprise organizations is now in urgent demand across companies of all sizes.”

“Our new release is powerful yet intuitive enough that data professionals — including analysts at medium-sized and smaller organizations — can now harness the power of AI to quickly join multiple datasets, generate millions of hypotheses and create predictive models, unearthing unexpected drivers for better decision-making.”