The Rare Find: Reinventing Recruiting
In late 2006, 22-year-old Adam D’Angelo confronted a serious problem. Facebook, then a small Silicon Valley startup, had picked him to be its chief technology officer. He was bursting with ideas about how to make the social-networking site bigger, faster, and more appealing. To make those dreams come true, Facebook relied on a couple dozen scruffy young engineers, crammed together in a graffiti-covered office. Reinforcements were desperately needed. D’Angelo and his colleagues refused to settle for any available programmer, fearing that lax standards would destroy the company’s innovative culture. Facebook was scrambling to master on-campus recruiting and to lure stars from Google and Microsoft, but those old-fashioned hiring channels weren’t paying off fast enough. Something quicker and more nimble was needed.
D’Angelo proposed that Facebook publish gnarly programming challenges and invite engineers anywhere to solve them. These wouldn’t be the superficial brainteasers that some companies used, like estimating the number of piano tuners in Chicago. Instead, Facebook’s website would issue multi-hour tests of coding prowess. With a bit of wit, these puzzles would find and deliver the right kind of people to the California startup.
Facebook engineer Yishan Wong volunteered to draft puzzles so hard that he couldn’t solve them. Before long, Wong and D’Angelo realized that their whimsy might serve a bigger purpose, too. “We developed this theory that occasionally there were these brilliant people out there who hadn’t found their way to Silicon Valley,” Wong recalled. “They might be languishing in ordinary tech jobs. We needed a way to surface them.” Goofy puzzles—some involved dinosaurs or gamblers—looked like the perfect bait.
Meanwhile, in Portland, Me., on a cobblestoned road known as Milk St., Portland Webworks was cranking out corporate websites for lawyers and consultants. Among the programmers there was Evan Priestley, a large, round-faced fellow in his early twenties. He was building expertise in Web applications development while growing tired of humdrum tasks—such as adjusting shades of blue for individual customers who weren’t sure what they wanted. “They were always saying: ‘Can you build us some Internets?’” Priestley later remarked.
No big-league recruiter was likely to rescue Priestley. “I had a pretty terrible résumé,” he later observed. He had quit high school a few weeks before graduation because classes became unbearably tedious. He switched majors three times at the University of Southern Maine for similar reasons. Eventually he left college without a degree. Most people pegged Priestley as a slacker.
One afternoon, Priestley had finished early, so he started reading an online news site, reddit.com. One posting alluded to Facebook’s puzzlers. Welcoming the mental workout, Priestley began wrestling with ways of automatically seating a clique of people in a movie theater, given that best friends want to be side by side and rivals need to be far apart. The puzzle looked hard and shapeless at first. After 45 minutes, Priestley cracked it. He double-checked his programming solution, decided it worked, and e-mailed it 2,500 miles west, to Facebook headquarters.
Impressed with Priestley’s approach, Facebook flew him to Palo Alto for a job interview. Engineer Marc Kwiatkowski tested the newcomer, face-to-face, on a trickier problem. As Priestley later recalled, “I told Marc what answer he probably wanted—and I explained why it was a badly constructed problem. You were supposed to speed up one piece of the code. But it didn’t address the fact that 98percent of the time was being wasted on network requests.”
Kwiatkowski smiled, and then decided Priestley was right. A week later, Facebook hired the man from Maine.
A new era of talent hunting has begun. It’s happening not only at high-tech companies such as Facebook, but also at Army bases, ad agencies, investment banks, Hollywood studies, corporate boardrooms, college admissions offices, and even at nanny agencies. In all these fields, experts don’t just sort résumés. They pick people and build teams in a profoundly different way. Traditional measures of past achievement, such as test scores and academic degrees, are losing power, and companies are getting better at looking for those future superstars who deliver many times the value of someone who is merely good.
At Google, Todd Carlisle is director of staffing. In 2005, he began an experiment, collecting factors that might distinguish between hiring great employees and picking the wrong people. Within a few months, Carlisle had some answers. Most supposed markers of success turned out to be mirages. When all the number-crunching was done, several dozen factors—including tidbits like the age when a recruit got into computers—emerged as ones that could help predict candidates’ chances.
What Google learned was that it had been looking at résumés far too narrowly. The company had started out by focusing inordinately on candidates’ education, grade-point averages, and even SAT scores. The thinking was that high-IQ people would do best at Google, and that the best way of gauging brainpower was to look at classroom records. Google ended up with lots of PhD holders from Stanford, MIT, Caltech, and top Ivy League schools. But by the time Carlisle got to work, Google was finding that some of these geniuses weren’t quite as effective as it had hoped. Even more important, company insiders worried that they might be turning away a lot of talented people whose true abilities surpassed their academic credentials.
“Take the wide view” became the overriding lesson of Carlisle’s experiment. There was room at Google for people whose grades had faltered because they were working 30 hours a week to pay for college. There was room for highly competitive people who had chased an athletic dream when they were younger—and now were applying that same relentless energy to professional goals. There was room, especially in nonengineering fields, for people who weren’t great students but had been running businesses, tutoring, volunteering, and otherwise being civic leaders from their teenage days onward.
Such candidates would stay invisible if Google rigidly scanned résumés the traditional way, from top to bottom; as many as 75,000 a week streamed into Google’s offices. So many candidates chased so few openings that if reviewers didn’t see some “Wow!” factor right away, they hit the “Reject” button. The best hope of spotting these hidden winners, Carlisle came to believe, was to steal a quick peek at the bottom of the résumé. He became known as the man who analyzed them “upside down.” Now, when Carlisle pulls one up on his laptop—which happens dozens of times a day—he begins by tapping the “Page Down” key a couple times until he reaches the final entries.
Then he scrutinizes the loose ends of candidates’ bios. “I want to know their stories,” Carlisle explained one morning. “I want to know what these people are all about.” In a moment, he might start hunting for the classic markers of competence: work history, education, credentials, and the like. But first he wants to see if some special, rare attribute could point the way to greatness.
Far more than the search for programmers and tech execs has changed. The U.S. Army’s Special Forces size up candidates in a 19-day selection process in which no bullets are fired. The Army isn’t looking for the world’s best shooters; it can teach that. What it wants are resilient soldiers who can tackle unpredictable challenges. One test is to ask a group of prospects to move a trailer that’s missing a wheel 2.5 miles. The soldiers have some pipes, rope, and other objects at their disposal. The bare axle can never touch the ground. What they don’t know is that there isn’t a good way to put the elements together to get it done. A sergeant watches not to find who has the best solution but who best handles the fact that there is no answer. The winning candidates may not be honors graduates of West Point, but they have the grit that can pay off in tough, isolated assignments overseas. Likewise, at a test facility known as the Shooting House, at the FBI’s main training site in Quantico, Va., recruits are watched as much for when they don’t fire as for when they do.
The art of talent scouting also involves knowing what shortcomings don’t matter, which flaws can be overlooked. In sports, for instance, where talent spotting has been developed into a high art, it often means discounting size, strength, or an odd delivery. Baseball scout Herb Raybourn made the best call of his career many years ago in a small town in Panama, while watching a 21-year-old converted shortstop trying to make it as a pitcher. The prospect’s fastball clocked in at an undistinguished 84 to 87 mph. But Raybourn loved the pitcher’s loose, easy motion—and saw lots of room for improvement. For a base payment of just $2,500 cash, Raybourn signed future Hall of Famer Mariano Rivera to a contract with the New York Yankees. Likewise, many of the scouts who first saw Tim Lincecum pitch in college knew that he had first-pick statistics, but couldn’t get over the fact that he was short and slight for a pitcher. The San Francisco Giants picked him tenth, and he won a Cy Young Award two years later.
Corporate directors are also taking fresh looks at the process of picking chief executive officers. Recent academic studies show that charisma and affability may be overrated as traits that lead to CEO success. Efficiency, problem-solving, and hard-nosed accountability appear to be more valuable. If directors aren’t confident they can pick out the winning elements, consultants such as Randy Street are devising new interviewing techniques and simulations meant to establish which highfliers deserve a shot. He spends four hours at a minimum discussing each candidate’s life. He is looking for details that can’t appear on a résumé.
Priestley’s first few months at Facebook were impressive. He started on a team of programmers figuring out how to speed up Facebook’s underlying computer infrastructure. Then he helped develop a framework that made it easier for users to add games, maps, and other applications to their Facebook experience. He was a computing polyglot, switching from PHP to Ajax to other programming languages as needed. He worked days; he worked nights. Whatever Priestley didn’t know, he learned in self-run tutorials. One time, Facebook’s site stopped working for a small group of users. It turned out their computers were burdened with an obscure, out-of-date security program. No one at Facebook knew what to do. At 11p.m., Priestley rustled up the Internet’s only publicly available information about this security program—written in Dutch. By daybreak, Priestley and a partner had gained familiarity with Dutch terms such as foutmelding and beveiliging. They rejiggered Facebook’s internal code, and the problem was fixed.
Priestley wasn’t the only genius from nowhere to arrive on Facebook’s doorstep. As the company’s puzzles became well-known, programmers spent as much as 40 hours trying to devise solutions. Most of their efforts didn’t work. About 10 percent of the submissions amounted to accurate, viable programs. Within that pool, a much smaller number of the solutions displayed true elegance. The standout candidates earned job interviews at Facebook. Contestants who passed all regular screening tests were invited onboard.
Among these new hires was Jonathan Hsu, a graduate of Harvey Mudd College in southern California. He joined Facebook’s engineering team in July 2007 and began working on online ad software. Soon a beguiling extra job opened up. The original puzzle creator, Yishan Wong, left for paternity leave. Facebook needed someone new to oversee its growing collection of software quizzes. At a minimum, that meant sizing up candidates’ attempted solutions.
Hsu became the Puzzle Master. Each time Hsu posted a new puzzle, he added mischievous details that typified Facebook’s jaunty culture. In one problem, contestants were asked to help identify intoxicated users who couldn’t type properly anymore. It was repackaged as the “Breathalyzer.” To deal with this imaginary scenario, contestants needed to catch flurries of typos in unfamiliar texts, while absolving passages with rarer errors. Solving this challenge required shrewd use of a sophisticated look-up function. Hsu’s version made this task seem like a late-night dorm game.
Even the Puzzle Master’s identity and contact information became a baffler. The puzzle site’s opening picture never showed Hsu or any other individual. Instead, visitors saw an odd version of a lowercase f, consisting of many brightly colored rectangular slivers, all pressed together. Most viewers thought it was abstract art. The right ones realized that the f was a puzzle in its own right, written in Piet. Piet is a computer language in which all terms are represented as rectangles in the style of abstract artist Piet Mondrian. Thousands of people looked at the f each week. During a 30-month span, only 42 figured it out. The program yielded a secret e-mail address at Facebook, so solvers could alert the social networking company to their prowess.
From 2008 to 2010, outsiders e-mailed as many as 200 attempted puzzle answers a day to Facebook headquarters. It didn’t matter if the vast majority of submissions were clumsy or wrong. It was so easy, fast, and cheap to evaluate entries automatically—and then have Hsu take a closer look at the best submissions—that puzzle solving became a valuable tool in Facebook’s hunt for new engineers. Mainstream candidates were also encouraged to try the puzzles. Willingness to spend hours on a puzzle helped establish who really wanted to work at Facebook; clever solutions sharpened Facebook’s ability to spot the best coders. As of February 2011, Hsu estimated, some 118 engineers, or 20 percent of Facebook’s technical workforce at the time, had solved puzzles as part of their path into the company.
One of Hsu’s favorite discoveries was David Alves, a 2001 high school graduate who was still working his way through San José State in 2009. Alves’s college wasn’t known as a computer-studies citadel. Once Alves solved a puzzle and caught the eye of Facebook’s recruiters, however, the sidelights of his résumé commanded attention. He regularly earned top-10 recognition in programming contests pitting as many as 600 entrants against one another. He was president of San José State’s computer science club. Alves might have come from an obscure school, but he was ready to contribute.
These days, Facebook has a big, polished recruiting department, led by Lori Goler, a former Walt Disney and EBay executive. Many of the gung-ho engineers who championed offbeat programming challenges during Facebook’s coming of age, including Puzzle Master Hsu, have moved on; puzzles now are more tightly integrated into other recruiting methods. Last month, Facebook installed a new batch of puzzles, involving variables such as k and n, to replace the old bafflers involving dinosaurs and Red Bull.
Even so, Facebook keeps hunting for engineers without staring at résumés. At least five times a year it stages free-wheeling coding contests either at university campuses or Facebook headquarters. These events, known as hackathons or Hacker Cups, take on a carnival-like quality, as contestants race to complete programming challenges within a matter of hours. At one contest last year, Facebook offered a prize of $500 and a chance to meet company founder Mark Zuckerberg. The winning entry was a smartphone app called Airchalk that lets users wave their phones and create images on a whiteboard. It delighted Facebook’s judges so much that they decided the creators deserved even more. The contest champions, brothers Hani and Islam Sharabash of the University of Illinois, were hired on as summer interns.