How do you know you are hiring the right candidates? According to the Society for Human Resource Management, the average cost per hire is US$4,129, and it takes an average of 42 days to fill an open requisition. Imagine that amount being multiplied by all of your open positions over the course of the year. Then there is the issue on the other side of the coin—what if you rush to hire and you don’t hire the right person? A study found that 41% of employers estimated a single bad hire costs US$25,000, and 25% put the figure at US$50,000 or more.
What can companies do to reduce the cost per hire, as well as to reduce the chance of a bad hire? One approach is to use machine learning. Today, machine learning is already being used to make recruiting more efficient in three different areas:
- Application and résumé review: Screening résumés based on keywords, leveraging social data to identify candidates, and using online questionnaires.
- Pre-engagement: Deploying artificial intelligence (AI) assistants and chatbots to respond to candidate inquiries or schedule interviews.
- Talent sourcing: Narrowing top candidates from a large pool using key attributes.
How is machine learning doing this? Machine learning iteratively applies algorithmic analytical models to preprocessed data to uncover hidden patterns or trends that can be used to flag ideal résumés to review, predict the correct response to inquiries in the pre-engagement, or identify the best candidates for talent sourcing. While all of these areas can help reduce time and money spent to fill the position, there is one we believe can make the biggest impact to ensure you are hiring the right person: talent sourcing.
It is critical to eliminate the bias that could become inherent in machine learning.”
Will machines have better success finding the right candidates for your open positions than your recruiters? Ideally, a computer will find correlations and patterns that you would overlook, which would lead to increasingly higher-quality candidates. Here are some considerations if you are thinking about using machine learning to help with talent sourcing.
To leverage machine learning, you need to first define the variables on which to “train” the system. The variables you should consider will depend on your approach. Are you sourcing passive candidates—people who aren’t actually looking for a new job—or are you looking to narrow top candidates from a large pool of applicants? If you are doing the former, you might want to consider attributes such as how recently or frequently they have updated their LinkedIn profile, because this could indicate that they might start looking for a job or are already on the hunt. Or consider factors affecting current employer stability (such as mergers and acquisitions, layoffs, and stock fluctuations). You could also look at market indicators to help predict a downturn in a particular industry or company, which might create a plethora of available candidates, giving you an early advantage.
On the active recruiting side, imagine receiving hundreds, if not thousands, of applications for open positions. Here is where machine learning can help narrow the top candidates, depending on the trainability of your data. Do you have enough historical and relevant data on successful candidates or employees to train your system? It’s as though you are looking for a “mini-me” based on the profile of the “ideal” employee. The attributes you consider here will depend on the role, but one approach would be to reverse-engineer the best fit by looking at the attributes of successful employees in that role, such as their work experience, industry, and work product. Other attributes to consider would be the number of jobs they’ve had in the last five years, their tenure in each job, major and extracurricular college activities for a recent-college-graduate hire, or hobbies (competitive sports might be good for a sales role). You could also leverage machine learning to target candidates who have a higher probability of success based on prior recruiting strategies.
This article was originally published on Profit online (https://blogs.oracle.com/profit/).