A computational model to predict the organisational performance of startups in South African incubators
Chengalroyen, Jarryd Jermaine
There have been several changes to the global economy in recent history. These are due to numerous factors such as globalisation, advancement in technology, accelerated innovations, and changing trends in demographics. These changes have resulted in the need to improve levels of entrepreneurship. Entrepreneurship plays a crucial role in the improvement of economic growth and development. It also plays a vital role in facilitating poverty reduction, creating employment and structural changes. Entrepreneurship is a tool which can be utilised to improve living standards and general well-being. Failure rates for new businesses, however, are extremely high. The success of new businesses is a necessary factor to grow the economy. Business failures, particularly for new businesses, are a waste of valuable resources which could be used to grow the economy. Business incubators have been created in order to solve this problem. Incubators add value by combining the entrepreneurial drive of a startup with a plethora of resources usually not available to these under resourced startups. There have been several models developed to predict the success of startups. For this research, rather than measuring only success or failure, organisation performance was measured. This study creates a computational model, using machine learning, which will be able to predict the organisational performance of start-ups within incubators, based on specific factors. The organisational performance has been defined as a composite of both turnover and number of staff employed. In order to create the model, a literature review was performed, in which 15 factors were determined as being significant in terms of predicting organisational performance. This was used to create a survey, which was distributed to incubators. There were 103 respondents to this survey. When doing statistical analysis on the results of the 103 respondents, only five factors were found to have statistical significance - age, number of founders, capital rating, professional advisors and education level. Statistically, the predictability of the initial statistical model proved to be low at 23,8% for turnover and 25,4% for number of staff employed. Using the random forest machine learning algorithm, the predictability was improved to 35,92% for turnover predictability and 40,78% for number of staff employed.
A research report submitted to the Faculty of Commerce, Law and Management, University of the Witwatersrand, in partial fulfilment of the requirements for the degree of Master of Management in Entrepreneurship and New Venture Creation, 2018
Chengalroyen, Jarryd Jermaine (2018) A predictive computational model for organisational performance of technology start-ups, University of the Witwatersrand, Johannesburg, <http://hdl.handle.net/10539/28566>