Data Science Campus
The Office for National Statistics (ONS) Data Science Campus was set up to work at the frontier of data science and Artificial Intelligence (AI), to deliver research with impact and build capability across the UK public sector. We build skills and apply tools, methods and practices; creating insight to improve decision-making for public good. We work with UK and international partners, drawing on their expertise and resources, sharing the benefits of our education and research programs widely.
In March 2017, the Campus launched with a core of well-qualified professionals, recruiting mainly from industry and academia to strengthen ONS and government expertise in data science across the UK. The Campus has delivered a series of high profile data science projects providing insight into key policy themes and stakeholders across government, making the results and code widely available for others to use.
The objectives of the Campus are to (i) deliver data science projects for public good in collaboration with stakeholders in the UK and worldwide; (ii) grow data science skills and support the data science community; (iii) assess the value of non-traditional data sources and new technologies; and (iv) develop data science methods and apply them to provide decision makers with greater insight.
Data science for public good
Our research programme is delivering tangible results in 2018 that have been used by policy-makers and analysts across the UK and internationally. Examples include:
• Analysis of new big data sources on ships and goods to inform trade, transport and environmental policy in the UK. We have worked with Statistics Netherlands and assisted the UN Global Pulse Lab in Jakarta to use our findings and methods in their own work.
• Development of a re-usable interactive open source tool to enable the UK and other governments to monitor performance against Sustainable Development Goals (SDGs). This tool was published by ONS and reused by numerous international partners.
• Analysis of free-text manifest data collated by ferry operators to understand port-level trade flows. This project applies modern techniques of natural language processing (NLP) to process these datasets automatically and generate a classification of the type of goods being transported.
We now have a pipeline of innovative projects, including using street level image recognition to identify trees in urban areas, automated free-text analysis to identify trade at UK ports and assessing the ease of access to services across multiple transport modes.
Capability for innovation
Since its launch the Campus has grown to nearly 60 data scientists, academic managers, and business delivery support officers. In addition, two cohorts of apprentices have been trained in the Campus and are now contributing to official statistics in other parts of ONS.
New learning and development pathways in data science were created at a range of different levels from Apprenticeships to providing Doctoral and post-doctoral project support. A new Masters in Data Analytics for Government is being delivered by three universities and modules from this course are being provided to UK civil servants. By 2018 we have trained over 150 civil servants and are on track to help train over 500 Government Analysts in data science by 2021.
International and cross-sector partnerships
Data Science is most effective as a team sport, depending cross-sector working to maximise impact. Our partnerships with industry and academia helped to jump-start the operation quickly and the Campus draws on their vital expertise as we grow and develop. Agreements are in place with over a dozen universities and the UK’s Alan Turing Institute for data science and AI. These have led to several successful learning and research initiatives. Agreements with commercial organisations are in place. These have already led to some effective joint research and learning activities. For example we are working with Barclaycard to analyse their payments data to provide new insights into the modern economy.
Internationally the Campus is working with several National Statistics Institutes, including helping to develop data science capability in Rwanda and help implement the Institute of Statistics there implement the code developed by our research programme. Around 150 delegates from across the world who were attending a Data for Development Festival toured the Campus: feedback from this event was particularly positive, such as this Tweet:
“Think official statisticians can't innovate? Visit [email protected] [email protected] for amazing example of #datascience research and excellent practice of transforming #NSO from within through #datainnovation and capability building #data4devfest”
With the International Monetary Fund, we are leading work to make use of mobile payments data to create methods for monitoring the remittances SDG indicator
The reputation of the Campus has grown rapidly. We have shared our experience and learning at international events - including the UN World Data Forum - and with visitors to our Campus in Wales from statistical agencies from across the world, including Canada, Mexico, New Zealand, Australia, Indonesia, China, Singapore and Rwanda.
Based on the recognised success of the Campus to date, the ONS is planning to extend its operations, significantly expanding the learning opportunities offers and the increasing the scale and impact of the innovative research programme.
What Makes Your Project Innovative?
We work at the frontier of data science, machine learning and Artificial Intelligence to deliver impact for public good. We investigate big data sources and new methods that have not been used before in government to inform policy-makers with new insights, indicators and statistics. The novel data sets we have analysed for public good include street level images, ships’ transponder data, hand written manifests and social media posts.
The Campus has recruited data scientists with excellent skills and experience in data science, hitherto a rare commodity across the UK public sector. Projects are run in close collaboration with a range of partners: policy-makers, to understand the big questions; other national statistical institutions, to share knowledge and experience; subject matter experts and academic data scientists, to draw from their skills; and commercial enterprises, to provide new data sources.
What is the current status of your innovation?
In 2018, we will deliver at least 20 significant research outputs with policy impact. This will scale to over 30 next year and outputs from our research programme are being used by our stakeholders to support decision making. The Campus has also developed, from scratch, an extensive capability-building programme across the UK Government and Devolved Administrations, helping over 20 different departments and agencies. A new Masters Degree framework is being delivered in different universities, a PhD programme in collaboration with the Alan Turing Institute was launched and we have worked with Treasury, the Ministry of Justice, Department for the Environment, Farming and Rural Affairs and the Northern Ireland Statistics and Research Agency, in particular, to enhance data science skills among their analytical communities. We are on track to help train over 500 government analysts in data science by 2021.
Collaborations & Partnerships
We work with others to maximise our effectiveness and impact. We have collaboration agreements in place with industry, academic and international partners and we draw on their expertise as we grow and develop. Agreements are in place with over a dozen universities, UK’s Alan Turing Institute for data science and AI and industry partners. We have launched joint research programmes, co-funded PhD research programmes, accessed new data and delivered education programmes through these partnerships.
Users, Stakeholders & Beneficiaries
Our programmes are benefiting public policy makers in the UK and beyond. We have worked with stakeholders in over 20 government departments and agencies to enhance decision making and service delivery for the benefit of the UK population. Our outputs and code are freely available and have been reused widely. We have supported international agencies and governments and developed mutually beneficial partnerships with industry, charities and schools to enhance data science impacts and skills
Results, Outcomes & Impacts
Just 18 months after its formal launch, our research outputs are being put to use by others in the UK and abroad and we have strengthened UK government data science capability with a multi-disciplinary team of nearly 60 drawn from different sectors. We have worked with more than 20 UK agencies and collaborated with international partners. In 2018, we will deliver at least 20 significant research outputs. This will scale to over 30 in the next year.
Our early work, including that on Sustainable Development Goals, estimating calorie intake in the population, analysing port and shipping operations and social media analysis has been used by our stakeholders to support decision making. We have delivered a successful apprenticeship program which has supplied 13 trained data scientists for the benefit of the wider ONS organisation, introduced a new Masters degree framework and are training over 150 analysts across UK government in data science, aiming for over 500 to be trained by 2021.
Challenges and Failures
As a new concept within UK government, we were initially unsure of the scope for applying data science within government and the appetite of other organisations partnering with us. We were also uncertain of our ability to recruit the skills we needed but were overwhelmed with high quality applications, exciting opportunities to address important public policy issues and others wanting to work with us.
Our research projects are run in an agile way with a brief, early discovery phase to assess feasibility. Some are stopped early where relevant data are not available or initial analysis shows that the stakeholder need is unlikely to be met by the approach. We have learned from the projects have been ended on this basis and do not regard them as failures, rather as a necessary part of an innovative programme of work. Our challenge going forward in to mainsteam the work of the Campus to maximise the impact of what has been achieved for public good.
Conditions for Success
Among the key ingredients leading to the successful implementation of the new data Science Campus was (i) a clear drive plus support from the top of ONS to move at pace and to be different; (ii) a “greenfield” site with no legacy outputs or systems to distract from innovation; (iii) core funding from the UK Government, so did not have to focus on bringing in more income or having our focus determined by purely commercial considerations; (iv) our culture, location, working environment, variety of work and – most importantly – our mission for public good were conducive to recruiting new talent into ONS; and (v) we were asked to innovate continually in what we did and how we did it and were not over-encumbered from the outset by a focus on putting a formal governance framework in place.
These factors meant that we could experiment with different approaches, establish a track record of early success and be rapidly responsive to what we were learning from those early experiences.
We learned from Statistics Netherlands’ Center for Big Data Statistics which launched six months ahead of the ONS Data Science Campus. We have had considerable interest from other statistical agencies worldwide in our experience and have an active partnership with The Institute of Statistics Rwanda in setting up their equivalent facility and programmes.
We have also supported the UN Global Pulse Lab and shared experiences with agencies from as far afield as Canada, Mexico, New Zealand and the Republic of Korea. Within the UK we are helping several departments build capability in data science, based on our experience. We publish the results from our research projects. The methods and our open source code can be adapted for use across multiple domains and is made available openly, for free.
We learned the importance made some good early external hires mixed with some experienced people from within ONS who had the ability and enthusiasm to get things done. We needed to learn quickly and adapt our plans accordingly.
In a "start-up" environment, it was necessary to take some risks, change our minds about the best approaches and adapt our plans in the light of experience. It was important that we accepted that we did not have all the answers, so we reached out to partner organisations for help and inspiration: we weren’t afraid to copy what works elsewhere. We partnered with industry and academia to help jump-start our operation quickly and continue to draw on their expertise as we grow and develop.
There are a number of specific adaptations we have made in the light of learning from our early experience. Initially we spread the skilled people we had too thinly between separate projects, and our data scientists did not have all the support they needed to deliver end-to-end projects most effectively. We introduced new roles and stronger – but still light touch – project delivery structures. We have raised the level of our apprenticeship programme in the light of the high calibre of our first hires and will offer a degree-level scheme from 2019. To maximise the value of our university links we will offer more short-term PhD internships rather than extend the programme of co-funded PhDs.
A principle lesson was the level and types of support we needed to provide in order to increase the impact of our data science projects. As a result, many of project deliverables include reusable packages and open source code, training programmes and secondments from our team to help embed our innovation into mainstream activity.
Several agreements with commercial organisations are in place. These have already led to effective joint research and learning activities. For example, we have launched a project, with ONS staff working at Barclaycard, to investigate if detailed payments data held by the bank could provide more granular economic insights for local areas. We have also held joint learning events such as a “Hackathon” and a share and learn day for data scientists and apprentices from both organisations.