e-HR Adapts Recommendation Algorithms
In order to insure their success, startups specializing in recruitment are leaning on a technique that comes from e-commerce.
What do the matching site Tinder, the e-commerce giant Amazon, and certain startups specializing in recruitment such as Hunteed, AssessFirst, Kudoz, and Clustree all have in common? They all use an algorithm to generate recommendations that help people find their soulmate, their ideal product, or the perfect employee.
“A recommendation algorithm allows the user to find the product or service of their choice, without wasting time on fruitless research,” explains Aurélien Hervé, CTO of the e-HR startup Hunteed. This startup consists of a dozen employees and specializes in connecting companies and headhunters.
“The function of a recommendation algorithm is overall quite simple. Look at Amazon, who has been using this technique for a decade, as a model in this matter. The site watches the behavior of the user. If the user buys a band’s album, the algorithm knows that other people who bought this album have also bought certain other products…which are then directly suggested to the client. It’s good way to learn about preferences and to increase revenue. We do the same. Based on the headhunter’s searches, we bring up a list of relevant profiles.”
This opinion is shared by David Bernard. An occupational psychologist, he is now the head of AssessFirst, a company of 25 collaborators specializing in predictive recruitment based on the famous recommendation algorithms. For 2017, he expects double-digit growth. “The recruitment sector is conducive towards algorithms. It’s a story of finding a match between two needs. A vast majority of job searches are done on the Internet. Finding what meets the expectations of both companies and employees is based on quantitative data (salary, years of experience, mastery of software…). This is what makes it possible to create algorithms. In these algorithms, we can even implement qualitative variables and soft skills like the ability to negotiate or their understanding of the client. For this we administer a 10-minute long personality test to the candidates. But this is only one example among many others. Each company has its own recipe for creating and developing its algo,” explains the entrepreneur.
At AssessFirst, once the employee’s profile is defined by the personality test and the quantitative variables, the algorithm calculates a match rating corresponding to the expectations of the companies. “We use variable statistics – for those who are curious, Khi2 and Cramer’s V – to perform significance tests. With a match rating of 60% and higher, we recommend the profile to the client.” And more often than not, this works.
The BHV, which used AssessFirst services to recruit its crew of vendors and department supervisors, successfully brought its turnover rate down from 17% to 9% thanks to these precious algorithms. “To achieve such results, the company must be willing to give us the data. This is essential for calculating the match rating. So for us, the HRIS managers are precious spokespeople.”
When data scientists have all of the data in hand, it creates a significant advantage for the companies who can find a rare pearl in just a few clicks and greatly reduce recruitment costs and the number of unsuccessful interviews. But for some people, algorithms are not foolproof and don’t pose a threat of immediately replacing recruiters. And it wasn’t a technophobe who said this, but algorithm specialist Aurélien Hervé. “By definition, it is impossible to find the ideal candidate. Skills alone are not enough – even with a match rating of 100%, the candidate can’t be made for the company. Personality plays a big role. The algorithm can direct the best candidates and suggest a brief relevant list. And believe me, this is already a lot!”