The text of a company’s website helps predict the growth of that company. Dutch scientists of the Jheronimus Academy of Data Science (JADS) came to this conclusion during their research into the DNA for growth of companies. In their research, they examined specific combinations of word usage on the websites of thousands of SME firms and looked into the correlation with the growth of these companies. In doing so, the researchers used the data and insights of company information specialist Graydon.
“We examined sets of words that say more about the quality, customer-orientation and openness of a company,” says Dr. Daan Kolkman of JADS. “With the aid of natural language processing, we were able to identify that these words can be valuable to predict the growth of these companies.” For this scientific research study, Graydon made the company and website information of over 165,000 companies available. The data of more than 8,000 firms met the criteria for carrying out the study.
The JADS researchers looked at specific words in a number of categories. Is the company actually focused outwards or inwards? How pioneering or innovating is a company? Is the company customer-oriented? Is the company capable of staying ahead of its competitors? The unstructured text found on the sites of the companies involved is processed with the help of natural language processing into information with which a computer can make calculations. Subsequently, machine learning is implemented to look for the connection between the growth of companies and, among others, the text on their company sites.
“We were really encouraged by the outcome of this study,” says Chris Oatts, director of data operations at Graydon UK. “It shows that the DNA of growth is expressed in many different facets.” The outcome of this research study will not automatically be included in the Graydon score models as additional research is required to examine how this outcome can be applied to a larger company population.
Machine Learning as tool to predict growth
With this study, the data scientists of JADS have scientifically proven that better results can be achieved for predicting economic growth with machine learning that with more traditional techniques. With machine learning, it is not humans, but the computer that examines which connections exist between different values. Company information specialist Graydon is already partially using this technique in compiling their new score models. “The predictive value of machine learning has already proved itself, but it is good to see that, here too, there is a real scientific basis for it,” says Rob Veneboer, chief operating officer of Graydon. “Thanks to the collaboration between the business world and science, we have come to discover some innovative insights.”
Graydon made the company data and knowledge of their own data analysts available for the JADS study. “Thanks to the quality of the data and their expertise, we were able to achieve some nice results in just one year’s time. The first results of the study will soon be published, while other findings are indeed an invitation to examine further,” says Prof. Arjen van Witteloostuijn of JADS.
A follow-up is guaranteed to come, and more students have already come to Graydon with new research proposals. Veneboer added: “It is always fascinating to see how much value you can get out of data. If is immediately of use to us is less important. We especially want the UK, the Netherlands and Belgium to be the development ground to expand the expertise of future data analysts. This will not only benefit us, but our whole economy as well.”