Google Wrote An Equation For Deciding Which Engineers Should Get Promoted - Here's Why It Failed

Google is famous for making decisions based on crunching data. In most cases it works for the company, but using algorithms and insights isn't always the best approach.

That was the lesson Google's VP of People Analytics, Prasad Setty, shared in an address at the company's re:Work conference (we first spotted the video of his speech on Quartz).

Setty says when Google first formed the People Analytics team about seven years ago, its objective was to make sure the company's people decisions were based on data and analytics. It took that goal very literally.

"We wanted analytics to spit out our people decisions," Setty says.

Promotions are a big deal at Google, he explained. Twice a year, the company brings together its senior-most engineers from all over the world to form committees and mull over the huge stacks of engineering promotion nominations filed. In each cycle, thousands of Googlers get promoted to positions of higher responsibility. The whole process takes several days.

Because the People Analytics team wanted to help its "engineering brothers and sisters make these decisions more efficiently," they came up with a decision making model to decide which employees should get promoted:

Here's the formula they came up with:

Apparently, in the team's tests, their model showed 90% accuracy for 30% of promotion cases and it seemed to be reliable and stable across multiple cycles.

"We thought that these people who lived in the world of search and ads algorithms all day long would love this," Setty says. "

Turns out though, they didn't like it one bit. The hiring committees completely shunned People Analytics' algorithm.

"They didn't want to hide behind a black box," he says. "They wanted to own the decisions. They didn't want to use a model to do so."

Subsequently, the People Analytics team shifted its approach, realizing that it "should let people make people decisions."

People Analytics shouldn't be trying to make algorithms to replace people, Setty realized. Instead, it should be all about arming its executives with better, more relevant information so they were capable of making better decisions.

Google's promotion, hiring, and on-boarding processes all still rely on information rooted in research, but they no longer entertain the idea of letting algorithms replace people.

Watch Setty's full presentation here: