Employee Recruitment and Candidate Assessment

Why Is It Important to Have a Good Employee Recruitment and Candidate Assessment Model?

The demand for skilled jobs high and employees increasingly stay with  company for only a few years. Due to this , companies find it increasingly important to ensure skilled employees are set up for long-term success and they turn to machine learning programs. While ML won’t displace in-person interviews, they can point to candidates with the most potential. In short, a good ML model can help companies reduce time to hire while optimizing the candidate funnel.

What Internal Data Should I Have for a Good Employee Recruitment and Candidate Assessment Model?

A good employee recruitment model incorporates as much information about successful and unsuccessful hires, past and present. It should also use job performance metrics and any information that would typically be found on the candidate’s CV. Examples include lists of skills, previous positions, responsibilities, and studies.

What External Data is Essential for a Good Recruitment and Assessment Model?

The key to a good recruitment model requires good natural language processing (NLP) but a great model may require a variety of external data. The most useful datasets are ones that can enrich candidate data with LinkedIn and other professional profiles for cross validation, skill extraction, and general analysis. Of course, background check data can be used to make sure that you are focusing on the right profiles if the position requires it.

What External Data May Prove Useful for a Recruitment and Assessment Model?

Many tools can automatically extract candidates’ social media and online presences to focus on the ones with the right skill-sets, capabilities, and social demeanor. However, additional data may include competitive recruitment data and salary data to make sure the position matches salary expectations.

What Are the Main Challenges of This Use Case?

There are a few key challenges for the recruitment use case:

1. Legal: Any recruitment and candidate assessment program should comply with local legal requirements. This action may prevent creation of a “one-size-fits-all” enriched data set.
2. Entity Resolution: It’s not always easy to connect a person’s name to data about them at scale. Most companies overcome this challenge by collecting as much first-hand data as possible, from LinkedIn profiles and email addresses. These, in turn, provide better entity resolution, ensuring no one falls through the cracks.
3. Defining Success: Few success metrics can be defined on a company level, but many can be defined per role or group. It is essential, therefore, to have a good definition of metric success in order to build a good training set, much less a set of accurate automated heuristics.

Interesting Case Studies and Blogs to Look Into

Oskar Hurme: 10 Machine Learning use cases for HR
The SHRM Blog: Case Study: AI and Bias in Hiring Practices

Tangible Examples of Impact

Using [Google’s AI for HR] program, Openlogix was able to search through more than 30,000 applicants to make a new hire within 24 hours. Previously, it took the firm around four weeks to hire someone. A study conducted by Delloitte found that companies spend 52 days and $4,000 to fill one open position.

The Burn-In: Google announces release of potentially revolutionary recruitment tool