Discovering the Most Promising Ideas in a Crowdsourcing Platform for Product Development

Abstract

Obtaining and analyzing key customer and product information from various sources has become a top priority for major competitive companies who are striving to keep up with the digital and technological progress. From this point, the need for creating an idea crowdsourcing platform to collect ideas from different stakeholders has become a major component of a company’s digital transformation strategy. Today, companies resort to idea crowdsourcing platforms to discover novel ideas from the public, employees, and vendors that they can use in their product development processes. However, these platforms suffer from problems that are related to the voluminous and vast amount of data. Different large sets of data are being spurred in these platforms as time goes by that render them unbeneficial or useless.

The aim of this thesis is to propose a solution on how to discover the most promising ideas to match them to the strategic decisions of a business regarding resource allocation and product development roadmap. The thesis introduces a 2-stage filtering process that includes a prediction model using a Random Forest Classifier that predicts ideas most likely to be implemented and a resource allocation optimization model based on Integer Linear Programming that produces an optimal release plan for the predicted ideas. The model was tested using real data on an idea crowdsourcing platform that remains unnamed in the thesis due to confidentiality. Our prediction model has proved to be 93% accurate in predicting promising ideas and our release planning optimization problem results were found out to be 85% accurate in producing an optimal release plan for ideas.