#926 1956 · AT&T · Corporate management / telecommunications
AT&T tracked 422 young managers for 20 years to test its own promotion instincts
问题
Managers were promoted on a supervisor's gut read of leadership potential, never checked against what happened later
背景
By the mid-1950s AT&T was promoting thousands of college hires into management every year on the strength of a supervisor's subjective read of who 'had it' — poise, drive, the kind of impression that reads well in a review meeting. Nobody had ever tested whether that impression actually forecast who would succeed twenty years later.
Psychologist Douglas Bray proposed running 422 new male hires through a battery of business simulations — an in-basket exercise, a leaderless group discussion, a mock business game — scored by assessors who then sealed their ratings and told no one, including the men's own managers, so the predictions could not influence the outcome they were meant to predict.
换别人会怎么做
Keep promoting on supervisor recommendation and interview impression, the method every large employer already used and had never subjected to a check.
他们看到了什么
Bray realized nobody had tested the promotion instinct against any alternative — so instead of pitching a better interview, he ran a sealed, blind experiment that couldn't touch anyone's real career.
那一手
Every candidate spent three and a half days at an assessment center performing simulated management tasks — clearing a fictional overflowing inbox against the clock, running an unstructured group discussion with no assigned leader, negotiating a mock labor dispute — while trained observers rated actual performance on the tasks, not resumes or supervisors' opinions, and the ratings were locked away for the study's entire run.
为什么管用
Sealing the assessors' ratings from every manager in the men's chain of command was the load-bearing move: it meant the predictions could not become self-fulfilling by influencing who got coached or promoted, so when the ratings later matched real outcomes it was because the simulated tasks genuinely captured something the interview missed, not because the study had quietly rigged its own result.
值了多少
Tracked 20 years, the 1956 blind ratings forecast who reached the fourth management level, beating the company's own process at its own job.
什么时候会失灵
A simulation only predicts real performance if the fictional tasks resemble the actual job closely enough that the same underlying skills are exercised; a generic personality quiz measures nothing, and the multi-day cost of running trained assessors on live tasks is why most companies still default to the cheaper, weaker interview.
后来呢
AT&T rolled the method out company-wide by the early 1960s, and it was copied outright by IBM, Sears, Standard Oil of Ohio, and J.C. Penney within a decade; the assessment center, now standard at large employers worldwide, exists in its current form because AT&T ran a twenty-year experiment on its own hiring instincts rather than trust them.
资料来源
- [1]The Management Progress StudyResearchGate / Journal of Vocational Behavior, 1988researchgate.net
- [2]History of Assessment CentersApostrof International, 2021apostrof.international