How to land a job in an era of 'big data'
Hiring managers are now using software to screen for problem-solving skills.
Would you make a good corporate executive? Do you have the right mixture of creativity, confidence, and willingness to take risks? Guy Halfteck may know the answer. Technology companies, financial firms, and retail chains have hired Mr. Halfteck to look beyond résumés and pluck out top-notch talent. His secret? Observing how candidates hurl virtual water balloons at cartoon plants.
Halfteck and his team at Knack in Palo Alto, Calif., created Balloon Brigade, a game for the iPhone and iPad that asks players to defend a tower against increasingly difficult challenges. The game's cute graphics are just a ruse. Neuroscientists and data analysts designed each level to measure how people approach new problems, learn from mistakes, and juggle priorities.
Massive corporations such as Xerox, Royal Dutch Shell, and Microsoft now rely on this kind of "workforce science" to help make smarter hiring decisions. Twenty minutes with Balloon Brigade can reveal more about a candidate's potential than a traditional résumé and cover letter, according to Halfteck and the heap of evidence that backs up this new field of research. By asking human-resource departments to team up with computer algorithms, companies hope to not only recruit better employees, but also cut out unintentional biases that might prevent qualified candidates from getting ahead.
Pulling off this kind of data crunching requires huge amounts of information about each job applicant. Pegged Software in Baltimore helps hospitals hire staff members by scanning through résumés, public information, and its own specially designed aptitude test. While Knack disguises its exam with whimsical characters, Pegged's test appears to be much more traditional. There are logic puzzles and math questions – but don't be fooled. The test is not what it seems.
"We essentially give people a test with questions that don't have answers," says Michael Rosenbaum, chief executive of Pegged. "We might give a puzzle that doesn't have a solution, or ask someone applying for a housekeeping position a calculus question."
Mr. Rosenbaum is not that interested in a person's answer. He wants to see how applicants react. Do they try their best to answer the question? How quickly do they give up? Do they circle back to unanswered questions at the end of the test? Each response turns into a data point for Pegged's software. Individually, these shards of information are worthless, says Rosenbaum, but together they form a mosaic – a statistical image of each candidate. Pegged then compares this data to existing profiles of successful and unsuccessful employees.
If a candidate closely matches the attributes of someone who has thrived in the position before, the software flags that application for further review. If, however, the data suggests that the same candidate would be a better fit for a different position – a job opening that he or she didn't think to apply for – the system can automatically pass along the application to the right department.
Before it hired Pegged, Shady Grove Adventist Hospital in Rockville, Md., watched a quarter of its staff turn over each year. Regularly training new employees was expensive. The ebb and flow suggested that workers were not happy with the jobs they found themselves in, and hospital administrators worried that this negatively affected patients' experiences.
Once the hospital incorporated Pegged into its hiring decisions, staff turnover dropped by 58 percent. After a year of fine-tuning the software, it fell by 77 percent.
Xerox uses a similar system from the data scientists at Evolv in San Francisco to hire call center workers that are remain with the company for longer than the average employee hired without the program. Another San Francisco start-up, Gild, combs open-source software and online forums to find high-quality programmers for companies such as American Eagle Outfitters, Salesforce.com, and Microsoft.
These systems collect thousands of data points on each applicant. This extreme granularity allows the systems to pick out subtle patterns that humans might overlook. For example, a client of Pegged needed two nursing aides, one to work in a long-term care facility and another for the acute-care unit. While the job descriptions were identical, the data suggested that an ideal candidate for one position may be a bad fit for the other. A résumé "can tell you something, but it might not tell you what you think it tells you," says Rosenbaum. "If you apply to be a nurse's aide and you are a leader in a community organization, you are more likely to succeed in the nurse's aide job in the acute-care facility, but you're more likely to fail in the nurse's aide job in the long-term care facility." He says the difference may have to do with the slower tempo of a long-term facility. Ultimately, however, the system has no idea why people with certain profiles excel in some jobs but not in others. It's up to humans to analyze what the software discovers.
By focusing only on the data, computers do not fall into a number of well-documented human biases. In 2003, the National Bureau of Economic Research released a paper called "Are Emily and Greg more employable than Lakisha and Jamal?" The researchers sent out essentially identical résumés, but some had white-sounding names and others had traditionally African-American names. The résumés with white names received 50 percent more callbacks for interviews. A similar report from 2010 found that the physical appearance of corporate executives was closely linked to how much money they made each year, but had no connection to the performance of the companies they led.
While firms could theoretically ask human-resource software to take names and physical appearance into account, such a move would likely be illegal. American companies are allowed to require candidates to take pre-employment tests, according to the employment law firm Moody & Warner in Los Angeles, as long as they stick to nondiscriminatory questions, such as tests for physical skills, mental aptitude, and drug use.
For more on how technology intersects daily life, follow Chris on Twitter @venturenaut.