Flatfile began life as a solution to the simple problem of importing spreadsheets. Cofounders David Boskovic and Eric Crane were working at Envoy, a workplace management software company, and were struggling with importing large files for customers. They wanted to buy a tool for the task rather than building one, but they couldn’t find what they needed, they told VentureBeat in an interview. “We were shocked to find that there was absolutely nothing on the market that could help us get data from a spreadsheet into our product effectively,” said Crane. “There [were] all these different open source libraries and pieces and components, but there wasn’t anything that could go into that product.” And so necessity became the mother of invention.
The company said in a press release that Flatfile can automatically match 95% of imported columns, using a combination of machine learning and fuzzy matching. Users can upload data via CSV, XLS, or simply paste from the clipboard.
Flatfile is also meant to be secure, by processing data in the browser — client side instead of server side, in other words. Because of that, Flatfile has no way of accessing any of a customer’s data. (At some point, Flatfile is planning to add that server-side processing, with high encryption, to handle its unique file structures and types.)
The company wants to keep the complexity of the operation hidden, meaning it doesn’t want users wasting time and expertise messing with it. But that’s not to say customers won’t interact with Flatfile. In fact, that’s part of the tool by design. Flatfile is essentially industry-, market-, and vertical-agnostic — it wasn’t built for any specific type of data input. That’s an advantage from the outset, allowing the company to make inroads into any number of different fields. It already has more than 30 customers, despite spending just a few hundred dollars on advertising to date.
But going generic generally means sacrificing a certain level of accuracy and precision. Flatfile knows this and relies on customers to help improve the product over time. Flatfile can flag data that doesn’t seem to be formatted correctly — or perhaps a typo or a new type of data format that’s germane to a certain industry — and then ask the user about it. The user can make any corrections necessary. “And then we leverage that as sort of a training set for the machine learning program to essentially then say, ‘Okay, in the future if I see [the anomaly], I might recommend this particular value,'” said Crane. “And then the user can validate that’s correct, and then that’s additional training for that machine learning. So essentially, over time, we can pick up on these different types of patterns.”
“So it’s sort of a ranked system — a combination of algorithms plus basic machine learning, plus then reinforcement and scoring,” added Boskovic.
The team has planned some natural extensions of Flatfile. First up is data healing, so not only would data get into customers’ systems, it would be of the highest possible quality. The company’s ultimate goal is to reduce the importation of data to a one-click operation. These features will likely roll out gradually, as Flatfile builds “expertise” in various fields by processing customer data and improving its data sets. “The technical innovation comes from the audience,” said Crane.
Though the company was technically founded in 2018, the Flatfile team didn’t go full-time until January 2019, during the soft launch of the core product. They’ve now secured $2 million — led by Afore Capital, along with Founder Collective, Designer Fund, Liquid2, and Gradient Ventures — to scale the small company and build out those aforementioned new features.