When I first started the Data Science Fellowship project, I wasn’t sure whether my own expectations would align with the COVID-19 distorted reality. I was joining the ECC as a ‘Data Science Fellow’, with all the impostor syndrome that accompanies such a grandiose title. Additionally, it was my first time joining a large public-facing organisation, mid-way through a national lockdown nonetheless!
I hurriedly tidied my North London 1-bedroom apartment before my first video call. Luckily, I needn’t have worried as my colleagues at ECC were very welcoming. The first Virtual Pub event started off with a round of casual icebreakers for the new guy – “What did you study?”, “Which university did you go to?” and “Name the worst thing you have ever done”. Having read the principle on ‘Openness’ in the ECC Constitution, I ended up confessing my sins to a Teams video call of 15 strangers in my first week!
After being on boarded onto ECC’s IT systems, I was quickly brought up to speed and had many video calls introducing me to different ECC departments. I joined weekly huddle meetings with the hard-working D&A and JAM teams, witnessing first-hand the broad range of work that is delivered each week. I was also lucky enough to join ECC’s (infamous) Away Day and had the opportunity to learn from and be inspired by Paul McGee, ‘the SUMO guy’. Despite the limitations of a fully remote work environment, I felt like I had quickly found my (virtual) place within ECC.
From the start of the Fellowship, I was entrusted with the autonomy to shape the data science project to meaningfully contribute towards the ECC’s broader initiatives. This challenged me to apply a data science mindset within a consulting-style framework, ensuring that real-world conclusions could be drawn from any data analysis.
I reached out to many different teams within the ECC and strived to ensure that the rationale of key decisions made throughout the project was transparent to all stakeholders, including technical details such as the chosen algorithm and hyperparameter tuning process.
Grounded by the D&A team’s investigation into deprivation and the ECC’s continued efforts to support vulnerable JAM households, the Fellowship project sought to model and map job skills demand within Essex districts using topic modelling, a Natural Language Processing technique. The text descriptions of ~12,000 online jobs listings were cleaned and processed before being used as input into the model*. These jobs listings were sourced from a broad range of industries within each of Essex’s districts, including healthcare, tourism, construction, manufacturing and retail. The resulting heatmap visualisations gave interesting insight into varying concentrations of job skills demand across different districts, whilst radar plots were used for quick comparison of the districts’ characteristics.
By using online jobs listings from publicly-available data sources ECC would have ongoing access to an additional evidence base to better guide the provision of district-level skills training. This targeted skills training could help to empower JAM households, alleviating some of the related socio-economic and psychological factors caused by their JAM circumstances.
Throughout the Fellowship project, I was supported by the Technical Experts at the Catapult and the Domain Experts from ECC. This support enabled both quantitative and qualitative analysis to be progressed simultaneously, greatly enriching the real-world relevance of the project’s conclusions and recommendations. I’d like to say a special thank you to Nicola Mallett, Dawn Temme and Jevon Harper for their endless patience and kind encouragement over the past 3 months. Without them, this Fellowship project would not have been possible.
From my 3 months at ECC, I have personally deepened my appreciation for the ‘behind-the-scenes’ work that drives public policy. I have also witnessed first-hand the herculean planning and late-night collaborations necessary to ensure a policy’s intended impact is successfully delivered to its public stakeholders. I’m very glad that the Fellowship project was able to contribute to ECC’s broader initiatives and look forward to seeing future data science projects from the D&A team.
*“This Fellowship project collected publicly-available jobs data from 12,000 online job listings and applied topic modelling, a Natural Language Processing technique. ECC domain experts from different departments generously gave their time to join an interactive workshop, where they were asked to interpret these topics as job skills. These job skills were visualised across the districts of Essex and offer an additional evidence base to guide the ECC’s provision of job skills training.”
For a more technical overview of this data science project, please visit this article on the Catapult blog.