Adatao is the latest in a big wave of startups promising to make business intelligence easier and more accessible for the masses, but the company stands out in a few ways -- most notably, it tries to bridge the gap between non-technical business professionals and the data scientists that many companies employ to make sense of the massive flood of data pouring in.
The company, which announced a $13 million venture investment led by Andreessen-Horowitz this morning, starts its pitch in a similar way as many other new-breed business intelligence companies like Birst and Tidemark. Old business intelligence systems from the likes of Cognos and Hyperion are too expensive, too hard to use, and take too long to set up.
Like many of these other newcomers, Adatao supports natural language queries, although Adatao -- cofounded by ex-Googler Christopher Nguyen -- uses Google-esque auto-fill to help sift through the query-specific jargon. (For instance, a demonstration of air travel data required the user to enter queries like "show relationship between dayofweek and arrdelay.") Results are displayed on nice looking, easily understandable dashboards using what the company grandiosely calls its "beauty layer":
But the real trick comes when a business user wants help finding patterns in the data -- a task typically left to a data scientist.
Adatao lets business users collaborate with data scientists directly within the app. So, for instance, a business planner at an airline might want some help predicting better delays. He could reach out to the in-house data scientist from within the app -- the process is similar to real-time collaboration in Google Apps, where Nguyen was director of engineering.
The scientist can then use a language such as Python or R to run a more complicated query to figure out which factors have the strongest correlation with delays.
As it turns out, the strongest correlation for delay is departure time, which isn't exactly surprising, although weather has a much smaller correlation than you might expect.
The data scientist could then use this information to work with the businessperson to create other fast queries that could be run on the fly -- for instance, to predict how late a particular flight is going to be.
In other words, Adatao is not just about getting easier answers to common or pre-canned business questions. It's about helping people figure out what questions to ask in the first place.
The plumbing used to make such queries possible is worth noting: Adatao runs against data aggregated into a single gigantic table in a Hadoop database, and does all processing in-memory. The filght delay demonstration was conducted against a data set of historic flight information from 1988 through 2008 stored in Amazon Web services, consisting of 124 million rows and 29 columns. In the demonstration, it took less than two seconds. According to Nguyen, doing the same query using MapReduce would have taken an hour.
The company began selling its product last December, and already claims a handful of Fortune 100 companies as pilot customers, including one telecom (Nguyen wouldn't say who) that has used Adatao to shorten a particular product cycle from 6 months to one week.
Nguyen says the company is focusing primarily on these big customers, who typically are already using Hadoop and just want a better way to visualize and run queries against the data there. To reach smaller customers, the company will work with partners who deliver data services and have been "screaming for apps" to run on top of them.
Andreessen-Horowitz partner Peter Levine will join Adatao's board of directors as a result of the investment. It's notable that the firm has led a number of other early investments into companies with a similar business intelligence focus, including Tidemark and Good Data. They're all taking slightly different approaches, and solving slightly different problems, but the basic underlying goal is the same: Helping businesses see, understand, and use more the information they're collecting.
This story, "Adatao bridges the gap between data scientists and normal people" was originally published by CITEworld.