Startup Grepsr Wants to Make Data Collection Easy

Grepsr is one of the more intriguing startups I’ve come across in a while. It is a data extraction and web crawling service which you can use if you need to automate data collection from places on the web. Initially I thought this was simply a data scraping tool, but upon closer inspection ‘service’ does seem to be a more apt description as there is some human involvement from the Grepsr team as well.

I spoke to Amit Chowdhury, who explained that Grepsr is a just a two-man operation so far. The service, which launched back in October, aims to give users a simple interface to tell Grepsr which data on a webpage they want. This is done with a sort of screencapture tool, where you can highlight or ‘snap’ the data that you want on a given webpage or document. There are also browser plugins that users can use to highlight target data as they browse websites (see picture). Once Grepsr has your requirements, along with any comments you might want to send them, they then create an extractor. Amit explains:

There is some manual intervention in this process, but we shield the users from this. It differs from [other solutions] in such a way that our focus is mostly on non-technical customers or companies who do not know how to go about data extraction… Grepsr hides the details, and just delivers the data neatly, organized and in a very streamlined fashion.

User dashboard

Browser Plugins

Amit estimates that the process is about 80 percent automated and 20 percent manual, the latter being the time needed to write an extractor for each project –though he points out that their “stable backend system makes it very easy and quick to write extractors.” Once an extractor is prepared, it can then be scheduled to collect data or update at certain intervals.

And Grepsr appears to present a number of options for data delivery, such as integration with Dropbox, FTP, and Google Docs, or you’re a developer they’ll provide data feeds that an application could consume. From what I can see by browsing a sample project, the results look very good.