Arloesiadur: An innovation dashboard for Wales
New data for innovation policy in Wales
Arloesiadur (innovation directory in Welsh) is a website that maps innovation in Wales, developed in a collaboration between Nesta and Welsh Government. The project is born from the belief that new data, analysed and presented in new ways can create valuable information for innovation policymakers. This includes web sources, such as company websites and social media platforms that are richer and more timely than official data sources, machine learning and natural language processing techniques that can generate predictions as well as descriptions, and help us understand reams of text, and interactive data visualisations and open datasets that empower users to answer their own questions, and carry out their own analysis.
Arloesiadur is an experiment to test the value of these new approaches: we scoped the needs of Welsh innovation policymakers, carried out agile pilots to explore the potential of different datasets - including official datasets about industrial activity, open data about research and web data about tech networking - and now present the results in the interactive visualisations in this website, developed together with infogr8, a visual content agency.
Here are some of our results, illustrated with visualisations taken from the website - click on them to explore them in the Arloesiadur website.
Findings: Responding to Wales’ economic challenges with innovation
Our work is motivated by the fact that Wales faces significant economic challenges such as low GDP per capita and productivity, which can only be addressed with a strong push for innovation manifested in two key strategies - Innovation Wales (2014), and Science Wales (2012) - and a myriad policy interventions from diverse bodies, ranging from Wales participation in MIT’s REAP programme, Cardiff’s City Deal and the Swansea Bay City Deal, the Office for National Statistics (ONS) Data Science Campus, and a myriad of initiatives to drive digital innovation and entrepreneurship, including the work of the Alacrity Foundation in Newport.
Our analysis of economic data highlights that Wales is already competitive in several high knowledge, high value-added manufacturing industries like aerospace or instruments, as well as sectors related to the green economy such as energy and environmental services (including recycling and collection and treatment of hazardous waste) which should gain importance in tomorrow’s sustainable economy. We have identified a corridor of manufacturing activity running from the North East to the Severn Bridge that added almost 7,000 jobs between 2010 and 2015 (an increase of 12.5 per cent).
Wales is also developing new strengths in knowledge-intensive and creative sectors such as R&D, creative services, computing and knowledge-intensive business services - however, most of this activity seems to involve small entrepreneurial organisations rather than larger employers. In 2015, businesses in these sectors represented 6.5 per cent of Welsh businesses (20 per cent more than in 2010), but only 1.5 per cent of Welsh employment. How can the creative and digital startups currently proliferating across Wales, be supported to become ‘scale ups’, and what connections could be formed between them and other sectors, including successful manufacturers further up north, to speed up the adoption of Industry 4.0 technologies (including Internet of Things, additive manufacturing and cyber-secure infrastructures) in Wales?
Our analysis of Gateway to Research (GtR) - an open dataset about publicly funded research, and web data from Meetup.com - a platform used by tech communities to organise networking and skills-sharing events, support the idea that Wales has the knowledge base and innovation communities to deliver this promise. In particular, our analysis of research trends suggests growing strengths in research areas related to the data revolution such as robotics and cybernetics, prosthetics, robotics and health, bioinformatics, statistics and data analysis, and security. In 2015 and 2016, projects led by Welsh organisations were awarded almost £5m by research councils. Some examples include uses of deep learning for cell imaging, led by Swansea University, development of robots that learn through play in Aberystwyth University, and a network to enhance big data analyses for plant research in Cardiff University.
Meetup activity in topics related to the data revolution, including big data, data science and machine learning are also an area where Wales is particularly competitive, with eight active groups and 1,741 participants in 264 data-related events, and attendees from all over the UK participating in activities in Wales. People have for a while been talking about a cluster of data innovation in the south of Wales, and our analysis suggests that this cluster is a reality.
A perennial challenge for innovation policymakers is dealing with ‘innovation system failures’ - the fact that people with complementary capabilities (e.g. businesses with a problem and researchers with a solution, or technologists whose knowledge could be combined into new products and services) will not necessarily talk to, or even know, each other. Although our analysis indicates that there is connectivity and cross-pollination between different research institutions and tech communities in Wales, with 283 research organisations in Wales involved in projects with other Welsh organisations in the last three years (45 per cent more than in the three years before), organisations still tend to look for collaborators close-by: a third of the research collaborations we identified were inside the same principal area; we also see limited participation in research collaborations by manufacturing businesses in the north west of Wales, in spite of that area’s industrial strengths. We hope that our interactive visualisations, by increasing the visibility of the research activities already taking place, might encourage even more and better collaborations going forward.
An unanswered question for Wales (and many other economies in the UK and beyond) is how to develop an innovative, knowledge-intensive economy, that generates benefits for all, instead of a ‘hourglass shaped’ economy with a small number of people in well paid, productive and innovative jobs, and the majority of people in low paid, routine jobs. Our analysis of the situation in Wales suggests that this could be a problem. The biggest employment segment in Wales is in the lowest rung of salary, and this segment has grown by five per cent since 2011. In Cardiff, 41,000 people - almost 20 per cent of the workforce - worked in industries in the top 30 per cent of median salary. This is less than half as many as those who work in the bottom 30 per cent (89,000 people).
Our predictive analysis suggests that Cardiff will gain business specialisation in some top paying jobs and some bottom paying ones, suggesting that its economy is becoming more unequal. Again, this raises important policy questions: what interventions (including access to finance and skills policy) can help balance Wales’ economy towards a less unequal structure, and what could be done to improve the productivity - and salaries - of important, low-paying employers like logistics, residential services or consumer retail?
Experimenting, iterating and sharing
Our work so far suggests that new data sources and methods can generate useful information for innovation policy. Now we need to monitor how they are used by policymakers, and identify the processes, skillsets and policy instruments that have to be in place to augment their impact.
We also need to reflect on the lessons from our research: for example, one of our data pilots, where we were using company website data to ‘snowball’ a network of innovative companies, did not work out within the timeframe of the project. Another one, where we have ‘scraped’ the websites of Welsh businesses and analysed their content to understand what they do, has not delivered results in time to be included in the interactive data visualisations (although we are making the data available here).
We are open about these things so that others learn from our experiments. It is also important to note that the website, as we are launching it today, is not ‘finished’. We will refresh the data in due course, and look for ways to augment it with new data sources and visualisations in the future. We are making as much code and data open as possible so you can peer-review what we have done, and build on it.
We look forward to hear what you think.