Medway Youth Trust is a registered charity serving vulnerable youth in the Medway area of England, approximately 35 miles southeast of London. At Medway Youth Trust, trained advisors provide advice and guidance to at-risk youth in the hopes of helping them to create brighter futures. MYT has pioneered the use of predictive analytics within the public sector; this work culminated in a Gartner Business Intelligence Excellence award for MYT in February 2012.
In this video, Gary Seaman, Head of Business Analytics for MYT, discusses how MYT is using Tableau to help advisors identify at-risk youth and allocate resources.
Tableau: What made Medway Youth Trust decide to try to Tableau?
Gary Seaman, Head of Business Analytics for MYT: I've always wanted to be able to visually show the output from the prediction. It's very difficult to explain prediction modeling to people who don't use data. So I wanted to show, you know, a clear visual way, particularly for key stakeholders
Tableau: Where do you get the data that you use?
Gary: Most of the information comes from the schools' systems—that gets uploaded once a year. We also have information that we need to pass onto the Department of Education on a monthly basis, the same as every local authority does.
And we have things called action plans. Our advisors will talk to a young person during an interview and ask them about their potential career choices, what they'd like to do, when they leave statutory education into further education. So all that's documented and we use a lot of that as well.
Tableau: So do you work with mostly structured data or unstructured data?
Gary: Both. Proportionately the structured data is about 20 percent. So it’s about 80% unstructured data. We do specialize in text analytics.
A lot of the unstructured information is just free text form notes that the advisor types up. It could be a young person would say that they're sofa surfing, which is a UK term for staying around at different friends' houses every night. And for me that immediately says they're at risk of becoming homeless. So we extract all the text from that, put it into a text category, and that's an identifier that we can actually use to see, “this young person is at particular risk of becoming homeless. ” So we can then deploy an advisor to help them.
Essentially we extract sentiment from that: what do young people like to do? What do they think about? We use relationship and entity analytics within that as well—so who are their peers, who are they talking to. All kinds of really, really useful information that we've now been able to use in Tableau as well.
Tableau: It sounds like you can get a pretty clear view of the young people that you help.
Gary: The clever thing that we're able to do is to join up all other data sources within our modeling software and within Tableau. We can connect data sources from schools' management system, from college systems, which is individual learner record information. So we're able to kind of map the path of a young person from year seven when they're 12, 13 years of age, all the way through to maybe 19, 20. We are trying to get higher education data so we can map a young person's education throughout their whole life.
Tableau: How does Tableau fit into to your process?
Gary: So from start to finish an advisor would type in the notes, it would be extracted, and then we'd put it through the modeling process. And the output from that, we're able to put into a Tableau dashboard where we can visually see their risk of becoming unemployed—we have a propensity score. And local authority advisors can see exactly who they need to target.
Tableau: You mentioned that it was difficult to explain prediction modeling to non-data people? How has that changed with Tableau?