Building a Better Hiring Algorithm

shutterstock_148166213 (1)

To handle the daily flood of résumés, corporate HR relies on keywords to separate the proverbial wheat from the chaff (and keep the chaff, if you believe the critics). But that practice raises one big issue: a conventional keyword-based system may overlook people who can do the job better, but don’t present as well as someone with a polished, keyword-optimized résumé.

Are companies surrendering a competitive advantage by overlooking the not-so-obvious candidates? Are they finding the right people?

In order to solve that conundrum, more firms are relying on algorithms that perform better than brute keyword searches. But even then, highly advanced math faces a challenge in landing candidates with the right mix of skills and experience.

Know Your Company, Know Your Employees

Infor Talent Science has been in the hiring business since 2002. Now a $3 billion company with over 12,000 employees, Infor handles the hiring needs of major corporations. The company’s 2014 acquisition of PeopleAnswers augmented Infor’s Human Capital Management (HCM) product offering.

“Searching for keywords does not cut it when making a good hiring decision,” said Jill Strange, director of HCM and behavioral science at Infor. The company’s solution is a scale-based questionnaire that measures 39 distinct behavioral traits of each applicant. That data is directly compared to the hiring company’s inventory of traits. “We don’t rely on subjective opinion,” Strange said.

There is a legal reason behind the methodology. For an algorithm-based hiring scheme to work, it must demonstrate that it doesn’t discriminate against anyone by gender or race, Strange explained. In order to maintain validity, the algorithm must undergo periodic updates; recalibration can range from once every 18 months to every four years.

The matching algorithm is the same for any company, in any industry. “It’s related to job performance,” Strange added. “The more behavior that is similar to a top performer, the more likely (the hire) will be a top performer.”

The follow-on value of Infor’s technique is measured by the retention rate. “People who fit into a job are more likely to stay,” Strange noted. If you can cut turnover for the employer, you also cut the cost of hiring replacements. That cost savings is presented as the return on investment (ROI) for Infor’s clients.

“Is a Hot Dog a Sandwich?”

Yes, that question appears on an online screening quiz offered by StaffGeek. The question has no right or wrong answer; you just have to agree or disagree. It’s one of 20 or so questions that StaffGeek uses to develop a client profile, which it tries to match with a suitable employer. The company refers to this as “distinct native attributes” (DNA).

StaffGeek founders Matt Reed and Sean Boyce didn’t fit in with the corporate culture at their previous companies. Inspired by that sort of frustration, StaffGeek seeks to find a “good fit.” Office culture is a key measurement: “We need to make sure that what is important to the candidate is important to the employer,” Boyce said.

The questions seek to uncover preferences. Does dress code matter or not? Is it okay to have music in the office, or is quiet better? Is working from home preferable? Flexible work hours: yes or no?

“Ours is not a one-size fits all approach,” Boyce continued. Candidates aren’t measured by any one metric. Cross-referencing candidates and the “DNA” of an individual company yields a “match quotient,” an indicator of a good fit.

Reed cited a Gallup poll that suggested 70 percent of all Americans are disengaged on the job: “They are not invested. They really don’t care.” Spending 40 or more hours a week on a job that you don’t like? “It’s nonsense,” he added.

Music to My Ears

When it comes to filling a position, Gapjumpers finds its inspiration in classical music.

Forty years ago, most symphony orchestras in the United States were largely composed of men, noted Kedar Iyer, the company’s CEO and co-founder. Then hiring judges began having candidates audition from behind a screen. “The judges only heard the music they played,” Iyer said. The process weeded out unconscious bias, and gender parity emerged in the following decades.

Much like the musician playing a blind audition, a programmer writing code can be judged without revealing his or her identity. “The jobs of the 21st century have a digital output. They can be objectively measured,” Iyer said.

Gapjumpers gives candidates a problem to solve as part of its screening process. The only thing that matters is the client’s performance—names and identities are hidden, but output is rated and ranked. That information is presented to the hiring manager, Iyer said. Women make up almost 60 percent of hires as a result of the change.

Hiring managers don’t necessarily know exactly what they want in a candidate. Gapjumpers will do a “calibration session,” sitting down with a hiring manager to see what they really want after viewing two or three profiles. The algorithm is adjusted after that meeting. “The subjective is abstract. We show them the evidence and place the objective input back into the algorithm,” Iyer said. “Fairness is allowing everyone an opportunity to tackle a problem an employer has set for a job.”

There is a Needle in Every Haystack

Crowdstaffing also tries to evolve past the static keyword-search model by acting as a digital brokerage firm between companies and candidates. Firms will send their job requirements to Crowdstaffing, which in turn posts the job on its Website as well as other industry job boards and social networks. Crowdstaffing does the work of screening the responses before presenting the results to the employer.

Static keyword search still gets results, but can also result in “false positives” or overlooked candidates, said Ravdeep Sawhney, the company’s vice president of technology: “It really lacks depth to understand the true capabilities and character of individuals until real-world, direct contact was made. “

The firm “connects a global network of certified recruiting partners, and each one brings years of experience in staffing for different industries and job categories,” Sawhney said. “We train our recruiting partners to learn about the candidate as an individual, beyond the résumé. This gives us a tremendous advantage in collecting data on every candidate interviewed, allowing us to match their personality traits with that of the company and position.”

Crowdstaffing’s algorithm also looks at a company’s past engagement, interviews, performance and placement data. “This data-mix tells us a great deal and helps in predicting with a high level of accuracy the probability of a candidate getting placed and being a top performer,” Sawhney explained.

“Because we specialize in Contingent Staffing, we are able to collect data during every stage of the recruitment and hiring process,” Sawhney said. “The depth of data available to us on a candidate is ever-growing, and this allows our technology to continuously learn and predict with a high level of accuracy which candidates will excel at which companies and what roles.”