Modelling online job advertisement performance

Abstract

Attracting talented individuals has become one of the most important factors ensuring an organisation’s success and is often seen as a source of competitive advantage. At the same time demographic trends of today such as retirement of baby boomers and smaller amount of young workforce, make recruiting qualified candidates more difficult. On top of these trends, the extension of the Internet has led to the development of online job boards, which has led to rapid growth of information and access to multiple different job ads for job seekers. Apart from the challenges, the use of digital channels in recruitment marketing also offers an opportunity unlike ever before. Using digital channels enables accessing detailed data that was not once available due to the limitations of traditional marketing channels. Much of this data is however not utilized to its full potential.

This thesis is a case study conducted with real life job advertisement data from a Finnish job board. One goal of this thesis is to gain understanding of how different features of an online job ad influence its performance in terms of apply clicks to page views ratio. The research questions are: 1) Which kind of factors affect the attractiveness of online job advertisements in general? 2) Which statistical method is most suitable for building an explanatory model for job advertisement performance? 3) Which features are critical considering the performance of an online job advertisement on a job board? This study consists of two parts: a literature review and an empirical part. The literature review focuses on job advertisement research, and different statistical methods. The empirical part is a multiple regression analysis supported by t-tests with the case company’s job advertisement data.

Previous research has identified various factors affecting the attractiveness of online job advertisements. The factors can be divided into three major groups: 1) factors related to the content of the job advertisements, 2) factors related to aesthetic features of the job ads, and 3) individual differences of the potential applicants. Based on literature, most suitable statistical method for building an explanatory model is in this case multiple regression analysis, as in this study the objective is to explain one metric dependent variable (job ad performance) with multiple metric and non-metric variables (job ad features), which fits the characteristics of multiple regression.

Based on the statistical analyses, several job ad features were identified to have a significant impact on the apply clicks to page views ratio. Variables with positive impact included for example working-age population in the job location, number of links in the job description, and use of a banner image. Variables with negative impact, on the other hand, included e.g. job description length and using bulleted or numbered lists in the job description. The independent variables of the final regression model explained 8.4 percent (adjusted R2) of the variation of the dependent variable with a significance level of 0.001. All in all, the results imply that explaining or modelling job advertisement performance is difficult. The rather low level of explanation of the model implies that the remaining 90+ percent of the variation in the dependent variable stem from other factors not included in this model. Most of the independent variables of this study reflect the aesthetic properties of a job ad. However, based on previous research, many factors related to the content of a job ad and individual differences of applicants affect job ads’ attractiveness. To increase the accuracy of the model, further research is needed to find a way to include factors reflecting the content of a job ad in a numerical form to the model.