How to Make Data-Based Hiring Decisions


Typical resumé fodder may not be predictive of an employee’s performance.

Even for potential tech hires, the job interview is still a pretty analog affair. You sit down, palms sweaty with a forced smile, and try to match the bullet points you crammed into your head last night to whatever your interviewer is asking. Or maybe you’re out to lunch, trying to look like you’re eating while still keeping your mouth free to answer a question at any moment.

The folks over at NPR’s Planet Money (one of the best podcasts, period, for anyone who’s unfamiliar) put together a great episode last month about how new data analysis tools may finally put an end to the traditional job interview.

Evolve, the company profiled in the podcast, uses forced choice tests, meaning they force test takers to choose between two statements that seem equally true or untrue, to suss out an applicant’s true character. Then they compare those profiles with current employees’ test results -- the good employees and the poor ones -- to see how they line up.

The company also looks at metadata, such as how long applicants take to answer each question and what Web browser they use. (For the record, Chrome and Firefox users stay longer and perform better at call center jobs than those who use Internet Explorer or Safari).

What makes this a truly data-driven process is that the company lets data from the best and worst employees’ tests drive what they’re looking for rather than making presumptions about what a good employee’s profile should be.

What they’ve found is that the traditional things people include in resumés, cover letters and interviews don’t often line up with what actually makes a good employee. The company’s data show, for example, that a software writer’s education level has almost nothing to do with how well she’ll perform on the job. For collection agents, who you’d think should be pretty no-nonsense, it turns out creativity may be more important than persuasiveness.

What makes this a particularly interesting story for government is that the profile of a superior government employee -- or what we presume that profile ought to be absent data -- is pretty messy.

To make just a short list: They should be highly motivated to work even though it’s more difficult for the government to reward hard work with performance raises and bonuses than it is for the private sector. They should be patient, understanding that government work progresses more slowly with more oversight and more paperwork than in the private sector. And they should probably be ideologically committed to public service, so their spirits won’t be broken by pay freezes and government shutdowns.

Those are all traits that are difficult to assess based on an interview or an application, though. If government managers used more data from their best employees in the hiring process, they might be able to draw more of those best employees into the workforce.

Listen to the full podcast here