Medway Youth Trust uses text analytics to fight youth unemployment

Medway Youth Trust’s use of IBM’s SSPS predictive analytics to fight youth unemployment attracts government interest.

Text analytics is at the heart of a project to combat youth unemployment. It's one of 351 proposals to save public money, part of the Cabinet Office's Innovation Portal initiative.

Medway Youth Trust uses IBM’s SPSS predictive analytics software to identify young people who are at risk of becoming “not in employment, education or training,” or NEET.

Gary Seaman, data quality manager at the trust, said, “We’ve used our NEET prediction model to improve young people’s life chances.” Seaman conceived the idea in 2006 to extend SPSS beyond the usual roster of supermarkets, financial services firms and mobile phone companies to “young people likely to become NEET.”

It was in 2008 that he got the chance to go full steam ahead. The Medway Council set up in that year the Medway Youth Trust – under the leadership, from February 2009, of Graham Clewes, chief executive – to deliver the already existing “connexions” service dedicated to getting young people into work. SPSS, now owned by IBM, was harnessed to identify youth at risk of becoming NEET. “I was attracted by the text analytics side of the software,” Seaman said.

Medway Council is a unitary authority that includes the towns of Gillingham and Chatham. The closure of the Navy dockyard in 1984 presaged a decline in employment and a rise in social problems. “The dockers’ houses are now being filled up with people with low self-esteem,” was how Seaman summed things up.

He explained that the Medway Youth Trust database is accessed by personal advisers who are in contact with young people and write up interviews electronically. “The personal advisers were coming to me and saying they had too little time to write up their notes properly," he said. "But the information was there in the text. We’ve always had this data. We are just using it in a more intelligent way.”

With the help of the supplier’s consultancy, the trust built an industry-specific dictionary using the vocabulary of the young people interviewed. “For example,” Seaman said, ‘sofa surfing’ is a term they use that means, in effect, ‘homeless.’”

In September 2010, the trust fed the characteristics of the prior NEET group into the model and then added in year 11 (15-16 year olds) data from the 2010-11 cohort of roughly 3,500 youths. The text analysis gave them 732 names of young people with a 70%-plus likelihood of becoming NEET.

From October 2010 until now, the trust’s personal advisers have facilitated offers of education, employment or training to 82% of that group. Seaman estimated that previously, 40% to 55% of those identified as likely to become NEET “would have had no idea what to do and would have fallen off the radar.”

Before using predictive analytics, the trust had to manually search thousands of client records to identify those most at risk and score candidates individually. “We found there were sometimes inconsistencies between the written notes of staff and the client records, which created errors in actions taken. This was especially true as it can be quite arduous reading text when looking at thousands of records,” Seaman said.

The trust has spun off a not-for-profit company, Hidden Patterns, to extend the SPSS implementation. “The NEET prediction model was the first, but you could develop a Further Education model to stop people dropping out of college,” Seaman said, confirming that his company has been working with MidKent College in Gillingham. “The Further Education sector has a big drop-out rate in January; a lot of students don’t come back after the Christmas break. From the application form data, local authority data and individual learning record data, you can use the software to predict those who have a high likelihood of leaving.”

He also confirmed that the trust’s model has met with interest at Central London Connections, another youth services agency that includes seven London boroughs.

“In a world of less money, this targeted approach is better than the blanket one of a few years back,” Seaman said. He also stated that the cost in welfare benefits of a young person being NEET is £56,000 over a lifetime. He conceded that the SPSS software, consultancy and training was “a significant investment,” but stated that for every one pound of investment the return is £3,200.

Read more on Business intelligence and analytics