DataKind UK: using data science for good

profile picture Yoti 4 min read

We invest a lot of time and effort into supporting other socially minded organisations. As part of our efforts, we regularly invite those organisations to use our Park area for meetups, talks and workshops.

Last week we hosted DataKind UK and Global Witness who spoke to a large audience of friends and supporters about their work uncovering problems in the UK’s Companies Register. In this guest post, Suzy East from DataKind UK tells us a bit more about the project they presented.


Using data science for good

What do you get if you cross four do-gooding data scientists, a corruption fighting nonprofit and more than 10 million data points?

That’s exactly what the attendees at our recent meetup – kindly hosted by our friends at Yoti – came along to find out.

Despite unprecedented temperatures and the World Cup beckoning, more than 80 people showed up on Tuesday evening to hear DataKind UK volunteers talk about their seven month endeavour to uncover the hidden patterns in UK corporate ownership data.


Who are DataKind UK?

We’re a charity that uses data science for social good. We manage teams of pro bono data scientists and technical experts to deliver on projects with our nonprofit partners. We’ve been running for five years now and have a thriving community of volunteer data scientists who love to use their data skills for good.


It started with a DataDive

It all started back in 2016, when we first worked with the anti-corruption organisation, Global Witness. We spent a weekend of exploratory analysis with them, which became known as DataDives. A team of 50 volunteers unearthed a wealth of insights on UK company ownership from the Companies House open dataset. Namely, a worrying lack of data integrity.

For example, in the nationality field, people had found over 500 ways to say ‘British’, including ten people who identified as ‘Cornish’.

This and other findings were fed directly back to Companies House, and they changed the way they collected data as a result! Problem solved, right? If only it were that easy.


Uncovering patterns

After an initial look at the data, it was clear there was more work to do. So in 2017, Global Witness and DataKind embarked on a DataCorps project. On DataCorps, we work with a charity partner over 6-9 months to build a data science solution.

The aim was to take a full snapshot of Companies House ownership data and build a network graph mapping beneficial owners, registered addresses and other key identifiers. This way we could better explore and visualise the data to spot emerging patterns.


A sneak preview

Some of the key findings which we shared at our June meetup include:

  • 4,000 owners are listed under the age of 2 – including one who has yet to be born!
  • Over 40% of the beneficial owners of Scottish Limited Partnerships (SLPs) are either a national of a former-Soviet country or a company incorporated there – compared to just 0.1% of all Limited Companies.
  • 5 beneficial owners control more than 6,000 companies – might some of these individuals simply be stooges put in place by the real owners?


For more info on what we unearthed throughout the project, check out the report from Global Witness, due to be released in July 2018.

As well as sharing some results from the project, attendees got to find out what makes for a successful data for good project, and what it’s like to volunteer on one.

If you’re interested in finding out more, or volunteering with DataKind UK, please come along to our next meetup or sign up to our mailing list.


Suzy East
Project & Events Coordinator
DataKind UK