Recruiting the right person for a leadership role is a significant challenge. As Forbes contributor Ken Sundheim put it: “Hiring good people is hard. Hiring great people is brutally hard. However, it’s brutally necessary”. This is even more true for leadership roles. If you want the best people you have to decide what your company wants and needs. What does your future leader look like? What’s his or her skillset?
After deciding the aforementioned questions, how do you find these potentially great men and women? Research shows that jobs requiring a bachelor’s degree are posted online 4 out of 5 times. Does this mean online job-ads are the all-out choice for high-level recruiting? Unfortunately not. The German association of recruiters (BPM) estimates that approximately 50% of management staff are recruited through personnel recruiting – a number which can (obviously) be taken with a grain of salt but also hints that potential leaders are recruited from multiple sources. In this blog post Jan Martin Priebs, data quality analyst at Textkernel, explains how he used Jobfeed to analyse e-leadership skills in UK vacancies.
By Jan Martin Priebs
How to get the camel through the eye of the needle – or how to process large text corpora
We want to find out what a typical leader looks like. What experiences and skills should the person have? We could turn to shiny guides and articles which tell us how resilient and innovative a leader should be. But what are companies really looking for? Let’s assume a company publishes a vacancy on various job boards, the company website and newspapers. Looking at one vacancy, or even a dozen won’t help us much. But looking at and analysing hundreds or more vacancies is a time-consuming task.
To launch a large-scale analysis of vacancies you either need a tremendous amount of manpower (and hours), or a little bit of technical help. Textmining tools allow us to analyse textual data without requiring to read every vacancy one by one. One of these methods is co-occurrence analysis. This form of analysis aims to display the global context of a word within a corpus. Statistical analysis determines how close words are in terms of meaning. There are many possible ways to measure how ‘close’ words are to each other. In this case the Dice’s coefficient and a log-likelihood test were used.
It should be noted that the content of the corpus has to have some similarities. Putting hour-long political speeches in the same corpus as tweets would not be the best approach. In our case we are a going for a diverse selection of texts which share the same overall topic and format.
We’ve got the tools, we need the data
Now we need something to use our tools on. Luckily, Jobfeed by Textkernel provides us with all the data we need. Jobfeed lets us search through large amounts of vacancies which are posted every day. But just choosing random vacancies won’t answer our question. To limit our focus we can make use of external knowledge. Based on a theoretical framework about modern e-leadership developed for the European Union, e-leaders are supposed to be IT-, leadership- and business-savvy. A selection of keywords enriched with synonyms and knowledge from other researchers can be used to identify appropriate vacancies in Jobfeed. This works by connecting every keyword in a group by OR and to groups by AND. The search would look like this:
(leadership OR executive OR …) AND (business OR commerce OR …) AND (…) NOT (…)
The idea is that each field of expertise is present and represented by keywords. With the help of Jobfeed’s keyword search and these Boolean operators (AND, OR, NOT), we get vacancies which contain at least one word belonging to each category.
The tech-savvy leader
The results below are based on a corpus of around 1500 vacancies. Co-occurrence analysis allows us to see the context of a word. When we are looking at keywords like the ones below it is possible to see their four most important connections.
The highlighted keywords are concerned with things a person can do or knows and things which help to fulfil certain tasks. One clearly distinguishable group are adjectives like “strong”, “excellent” or “deep”. This is not surprising for obvious reasons. Another result is that keywords like knowledge, understanding and expertise are linked to terms from the respective field of work (e.g. “industry”, “technologies”, “technical”). There are also terms which focus more on human interaction like “communication” and “leadership”. The prominent occurrence of “communication” is supported in scientific leadership literature. This is because communicating intentions and goals is assumed to be one of the most important leadership skills.
The results can be summed up as the most common things companies are looking for. So our typical leader would look like this: in addition to a deep understanding of industry and technical knowledge, companies are looking for a top-notch communicator. A person who can apply his or her technical knowledge and communications skills in different ICT-supported environments effectively. Be aware that these results are, of course, just the surface and a small part which is highly dependent on your keywords and selection. There are also many more possible ways to use various textmining tools to exploit the riches of Jobfeed’s big data capabilities. This is just one, small example!
About the author
Jan is a data quality analyst in the Jobfeed team at Textkernel. His responsibilities include maintaining and improving the data quality for all Jobfeed countries. He graduated from the University of Bonn in 2016 and his background is in sociology, labor market research and political science. In his spare time he likes to read and watch science-fiction novels and movies, work out and go bouldering.
Are you curious about Textkernel? Check out our open opportunities at textkernel.careers.