Zooming in on small innovative companies in the Netherlands

‘Innovation is one of the primary sources of economic growth,’ says Henry van der Wiel of EZK. ‘It contributes to solutions for current societal challenges, for example in the field of environment. Innovation also boosts labour productivity and the earning capacity that can help pay for these solutions.’ Innovation is hence a priority for Dutch central government; it has been agreed that 2.5 percent of GDP must be spent on Research and Development (R&D). ‘In order to monitor policy goals as a government department, we require information,’ explains Van der Wiel. ‘We know that the data CBS provides on innovation do not include small companies, but they do form a significant part of the Dutch economy.’

Complementing each other

Both CBS and Innovatiespotter have covered the topic of innovation for years. The request from the ministry provided them with an opportunity to find out to what extent both organisations complement each other. ‘We each adopt our own approach,’ says CBS researcher Bob van den Berg. CBS works based on key registers and surveys as well as detailed company data it receives. Innovatiespotter uses, apart from machine learning technology, so-called ‘expert assessments’. CEO Gea Vellinga explains: ‘We are continuously retrieving new data on innovation, technology and sustainability from the entire enterprise population of the Netherlands with tools such as webscraping and machine learning, so we have an overview available of small and startup companies as well. In addition, we tend to look at economic sectors in a different way. Traditionally, companies are classified under one particular sector, but we take a different approach. In the energy sector, for instance, we find not only companies in the oil or wind energy business, but also in specialised transportation of wind turbine components or the installation of heat pumps.’ As part of the project, Innovatiespotter conducted a pilot study to chart the energy sector. Vellinga: ‘That pilot yielded new insights into the existing energy ecosystem. In the northern part of the Netherlands, there are around 1,500 companies actively offering energy-related products and services. A sizeable share counts as innovative.’

‘Nearly 30,000 innovative companies identified with a new method developed by CBS’

Learning model

Over the past few months, CBS has investigated how its scope can be broadened without adding to the survey response burden on companies. Van den Berg explains: ‘We now also focus on companies with less than ten employees. We have designed a model that identifies innovative companies with the aid of webscrapers and machine learning, based on their websites. This is a workable model, although the terminology describing innovation tends to evolve rapidly, which means that this model needs regular training.’ Innovatiespotter studied whether their approach could complement the existing research results produced by CBS. ‘Our results are timelier than at CBS, as registers and surveys are data on the past. Once we find new information on company activities with the help of our webscrapers, we move on to processing it immediately.’

Sharp picture

Collaboration between CBS and Innovatiespotter has gone well. Vellinga: ‘We receive daily updates on our dataset. CBS works with reference data as they need to be able to report on specific time periods for their statistics. Sometimes, this makes it difficult to draw a comparison. Another major difference is that CBS produces statistics and looks at averages. We work with overviews of companies. Our results must therefore be totally accurate. It is precisely these differences that have taught us a lot about each other’s work.’ Van den Berg explains: ‘We’ve achieved a good result by viewing our respective methods and results as complementary.’

Official statistics

Van der Wiel is content with the results: ’Close to 30,000 innovative companies were identified with the help of CBS’ new method. This number shows you that small companies do make a relevant contribution to innovation.’ The results are still experimental. The method still needs to be validated. ‘Besides, these are merely data on technological innovation, whereas innovation may equally take place in the social domain, for example. That is why we shall be happy to discuss future projects with CBS and Innovatiespotter. We’re not there yet, this is only the beginning.’ Van den Berg agrees: ‘This project has definitely not been concluded yet. We’re positive about it, but now the challenge is to turn the results into official statistics on small innovative companies.’