Do you have a hard-to-fill job? Not enough qualified applicants? You could search through a candidate database. Or, even better, save yourself the work and rely on algorithmic candidate recommendations. This is the message from a newly published paper in the Journal of Labor Economics by my economist colleague John Horton, from New York University’s Stern business school.
Algorithmic recommendations can increase the job filling rate by 20 percent
John collaborated with the ODesk (now Upwork) online work platform to do some experimentation in 2011. Employers were randomly assigned to a control group (business as usual) and to a treatment group (algorithmic recommendation). Without the recommendation, employers can either look at candidates who applied to their jobs, or search candidates on the ODesk database. With the recommendation, employers got up to six candidate recommendations based on their job opening. Employers with technical jobs (e.g., web programming and mobile development) typically had a hard time filling their vacancies, but the job filling rate was 20 percent higher for those who saw candidate recommendations.
Interestingly, candidate recommendations did not dissuade employers from hiring other non-recommended workers. This shows that candidate recommendations were not duplicating employer efforts, but were a real value added to employers.
Algorithmic recommendations are especially effective for jobs with few applicants
It is interesting that, in the ODesk experiment, algorithmic recommendations only had an effect for technical jobs and not for nontechnical jobs. When digging deeper into the data, John found that part of the difference could be explained by the fact that nontechnical jobs already had plenty of applicants. It is for the hard-to-fill technical jobs that the algorithmic recommendations were more effective, increasing job filling rates.
Recommended candidates are as good as employer-sourced candidates
When comparing recommended candidates and those sourced by the employer themselves, John found that recommended candidates were just as skilled and performed just as well on the job. Therefore, employers can rely on algorithmic recommendation to do at least as good of a job as themselves when sourcing candidates. This is especially remarkable given that the recommendation algorithm used at the time of the experiment (2011) was quite simple, and more recent algorithms are much more sophisticated.
For example, CareerBuilder now has a Talentstream match product that uses advanced natural language processing techniques to match jobs with the most suitable candidates. The product is currently available to staffing companies and other companies that use the Bullhorn ATS. The search engine allows employers to search not only in their candidate database but also across other sources like LinkedIn. Furthermore, search can be refined by specifying the relative importance of different skill requirements.
In conclusion, if you regularly find yourself with hard-to-fill jobs, you could benefit from algorithmic candidate recommendations.
Get in touch with your sales representative to learn more.
Get CareerBuilder’s expert recruiting tips and trends, right to your inbox.