How it all started…
Back in 2014, data scientists working in the Research Division at the global investment management firm, Winton Group, (Winton) were facing a problem. They wanted to test a specific research hypothesis, but couldn't find a vendor to sell them the necessary data. So they were left with a dilemma: forget about the investment idea or build the dataset themselves. They opted for the latter but it was an imposing challenge, requiring them to harvest data from news archives and regulatory filings over the last 50 years. To accomplish this, they came up with a software platform to help them break the problem down into manageable chunks of work.
And that’s how Hivemind was born.
They soon realised Hivemind could help across many of their projects. Much of their work involved dealing with complicated, unstructured documents: labelling them, collecting data from them, and using them to help verify or wrangle with the intricacies of existing data feeds. Many of these tasks could not be achieved to an acceptable standard purely computationally, and prior to Hivemind the highly-skilled team had to complete a large volume of tedious, often trivial, data tasks by hand. Hivemind made the team more cost-effective and allowed them to tackle projects which had previously seemed intractable.
For Winton—who had been pioneering a data-driven, systematic approach to investing in global markets for more than 20 years— Hivemind’s development proved transformational in its approach to data. By 2017, Winton had used Hivemind to build over 30 proprietary datasets from primary sources for use across their research and investment teams.
The platform’s success resulted in Hivemind’s incorporation as a stand-alone company in 2018 with the former Head of Research Data at Winton, Dan Mitchell, as its CEO. Hivemind now helps many more companies with their large-scale data collection, labelling, wrangling and elicitation challenges.