Google came up with a decision to decide the promotion of the algorithm, and then did not then ...

Google a little run off, tease me ~ practice a bit ~

Prasad Setty is vice president of the Google people Analytics team. The team that Google set up 7 years ago is responsible for collecting and using data to support the company's management practices. Its mission is simple, which is to make all personnel decisions based on data and analysis. At the Google Re:work Conference in October this year, Setty introduced some of the practices that the team uses to manage human resources in science. The conclusion is that, although the algorithm is good, can not be abused, personnel decisions must ultimately be decided by someone.

Google is a company founded by engineers and is still under the control of Engineers. The thousands of big companies make a lot of decisions every year: who should they recruit? Promotion who? How much salary should the best man pay? Usually Google will find 4, 5 senior engineers to form a committee, each committee review a pile of nominations, after many dialogues to make a decision. Google's move to the review process is rather complicated, with too many materials to review and meetings to hold, so that even Google's conference rooms are not enough, so go to a nearby Marriott hotel for a meeting.

As a result, in order to help mitigate the workload of the Review Committee, the People Analytics team developed an algorithm to streamline the decision-making process for staff promotion early on. The algorithm is a formula for calculating the possibility of promotion, as shown in the three factors that consider average performance, manager recommendation and personal recommendation (Google allows employees to self-recommend) (each gives different weights, the average performance weight is the most, followed by the manager recommendation, and finally the personal recommendation).

Comparing with the final promotion results, it is found that the algorithm is very reliable, the test results in the background are very good, and the performance is stable after too many cycles, of which 30% of the promotion case decision-making accuracy rate reached 90%. The team members were so excited that they were able to save 1/3 of the Committee's work, freeing them to devote their time to the most difficult decisions.

But the result is that the people do not buy, do not want to use this model. Because they do not want to hide behind the black box, but want to make their own decisions. So the algorithm has never been used to make promotion decisions.

Setty's lesson is that personnel decisions must be made by people. However, people Analytics can still work by helping decision makers with better information (using models to test their decision-making processes), but not using algorithms to make decisions for them.

And this insight has helped drive improvements in Google's human resources and management approach. People Analytics has fundamentally reshaped Google's recruiting system in many ways. For example, now Google has no emphasis on GPA (Gerschgorin Personality Assessment) and graduate school, but rather on a number of soft traits, such as "humility", "learning ability" and so on.

People Analytics also summarizes the 8 traits of the great manager through data analysis:

• 1) is a good coach
• 2) give the team authorization, do not do micro-management;
• 3) expressing interest/concern for the success and personal well-being of the team members;
• 4) efficient/result oriented;
• 5) Good communicators-listening and sharing;
• 6) Help team members to develop their careers;
• 7) have a clear vision/strategy for the team;
• 8) have important technical skills to help the team to provide advice in addition, Google is also looking for volunteers to conduct long-term research, set up a number of data points to track their decades of career performance, attitude, beliefs, problem-solving strategies, challenges and stress resistance. Although it has not been determined what can be found, collecting data research is certainly the first step in using scientific methods to study human resource issues.

Google came up with a decision to decide the promotion of the algorithm, and then did not then ...