The Numbers Game: How Turnover Data Becomes the Only Metric That Forces Institutional Accountability
Scott Rudin is a Hollywood producer. He is known for his work. He is also known as one of the nastiest bosses in the industry. The Wall Street Journal estimated that between 2000 and 2005, Rudin went through 250 personal assistants. Rudin disputed this number. He claimed his records showed only 119 assistants during that period — but admitted this estimate excluded assistants who lasted less than two weeks.
119 assistants. In five years. That is 23.8 assistants per year. That is one assistant every 2.5 weeks. Even using Rudin’s lower estimate. Even excluding the assistants who lasted less than two weeks.
The assistants who left told the Journal what happened. Rudin routinely swore and hollered at them. One was fired for bringing the wrong breakfast muffin — which Rudin did not recall but admitted was “entirely possible.” Another was called at 6:30 AM to remind Rudin to send flowers for Anjelica Huston’s birthday. At 11 AM that same day, Rudin called her into his office and screamed, “You asshole! You forgot to remind me to get flowers.”
The screaming is the problem. The abuse is the problem. The humiliation is the problem. But none of these things appear on a dashboard. None of these things force the institution to respond. What appears on the dashboard — what forces institutional attention — is the number. 119 assistants. Five years. The institution can ignore screaming. The institution cannot ignore turnover.
The core idea, stated plainly: Institutions do not respond to psychological harm. Institutions do not respond to moral injury. Institutions do not respond to the daily, cumulative damage of working for an abusive boss. Institutions respond to numbers. And the only number that consistently forces accountability is employee turnover. Not the abuse itself. Not the harm caused. The operational cost of replacing people who leave. Turnover is the only form of damage that is institutionally visible — because turnover is the only form of damage that institutions measure.**
What the Number Actually Measures.
119 assistants in five years. This is what Rudin admitted. This number measures something. But what it measures is not “how abusive Rudin is.” What it measures is “how many people could not endure working for Rudin.”
The distinction matters. The abuse happened to every assistant. All 119. Plus the ones who lasted less than two weeks and were not counted. Plus the ones who are still working for Rudin and have not yet left. The abuse is persistent, daily, cumulative. The abuse affects everyone in proximity to Rudin. But the number only counts the people who left.
Which means: the number undercounts the damage. The 119 assistants who left experienced abuse and then escaped. The assistants who stayed experienced abuse and are still experiencing it. The assistants who lasted less than two weeks experienced abuse so severe they left immediately — but they are not counted in Rudin’s official estimate. The turnover number captures only the exits. It does not capture the harm.
And yet: the turnover number is the only metric the institution cannot ignore. The institution can dismiss complaints. The institution can reframe abuse as “demanding standards” or “passionate leadership.” The institution can tell victims they are too sensitive, too dramatic, too unwilling to handle pressure. But the institution cannot dismiss 119 assistants in five years. That is not a complaint. That is a number. And numbers demand explanation.
The turnover metric is structurally incomplete. It measures exits, not harm. It counts the people who left, not the people who are suffering but staying. It captures the endpoint (departure) but not the process (daily abuse). But incomplete metrics are still metrics. And metrics — unlike complaints, unlike testimonies, unlike descriptions of psychological harm — force institutional accountability. Not because institutions care about harm. Because institutions care about what they can measure. And what they can measure, eventually, they must address.
What the Institution Sees vs. What the Institution Ignores.
The institution sees: 119 assistants in five years. This is a problem. This is operationally expensive. Recruiting costs. Training costs. Onboarding costs. Lost productivity during transitions. The institution can calculate these costs. The institution can put these costs on a spreadsheet. The institution can present these costs to leadership. And leadership — which may not care about assistants being screamed at — will care about the operational expense of replacing 119 people in five years.
The institution does not see: the assistant who is called at 6:30 AM. The assistant who is screamed at for forgetting to remind Rudin about flowers when the assistant was the one who reminded Rudin in the first place. The assistant who is fired for bringing the wrong muffin. The assistant who dreads going to work. The assistant who has panic attacks. The assistant who is developing health problems from stress. The assistant who stays because they need the job, because the credit looks good on a resume, because they hope it will get better, because they do not know what else to do.
This damage is real. This damage is severe. This damage is more widespread than the turnover number suggests. But this damage does not appear on any dashboard. There is no metric for “assistants who are psychologically harmed but have not yet quit.” There is no line item for “moral injury caused by daily humiliation.” There is no KPI for “employees whose health is deteriorating due to abusive supervision.”
The institution measures what is operationally expensive. Turnover is operationally expensive. Psychological harm is not — until it converts into turnover. And that gap — between what institutions measure and what actually harms people — is not accidental. It is a structural feature of how institutions define “damage.”
Why Abuse Is Tolerated Until It Becomes Operationally Expensive.
Rudin’s abuse did not start in year five. Rudin’s abuse started in year one. Probably earlier. The abuse was happening the entire time. But the institution tolerated it. Because the institution could tolerate it. Because the cost of tolerating the abuse (assistants being screamed at) was invisible, and the cost of stopping the abuse (confronting a powerful producer) was visible.
Then the turnover data became visible. 119 assistants. Five years. Suddenly, the cost of tolerating the abuse was measurable. Recruiting costs. Training costs. Onboarding costs. Reputational costs (the Wall Street Journal is writing about your turnover problem). The operational expense of replacing 119 people in five years exceeded the political cost of confronting Rudin.
But here is what did not change: the abuse itself. Rudin was not more abusive in year five than in year one. The assistants in year five were not more damaged than the assistants in year one. What changed was: the institution could no longer pretend the abuse was not a problem. Because the institution had a number. And the number was large enough that ignoring it was more expensive than addressing it.
This is the pattern across institutions. Abuse is tolerated until it becomes operationally expensive. Complaints are dismissed until they accumulate into turnover. Individual victims are told they are the problem — until enough victims leave that the institution can no longer staff the role. At which point the institution acts. Not because the abuse was wrong. Because the turnover was expensive.
The institution’s response to turnover is not moral. It is operational. The institution does not intervene because people are being harmed. The institution intervenes because replacing people is expensive. And this distinction matters — because it reveals what the institution actually values. The institution values operational efficiency. The institution does not value employees’ psychological safety, dignity, or well-being. Those things are not on the dashboard. Those things do not force accountability. Until they convert into a number the institution cannot ignore.
What This Looks Like in an Indian Office.
In Indian workplaces, the turnover metric operates with particular visibility in IT, call centers, and high-churn industries. The manager who screams at employees, who assigns impossible deadlines, who humiliates people in meetings — this manager is tolerated. The employees complain. HR knows. Leadership knows. But the manager delivers results. The manager meets targets. The manager is “tough but effective.”
Then the turnover data becomes visible. The team has 40% annual turnover. The company average is 20%. Suddenly, HR cannot ignore the number. Suddenly, leadership wants to know: why is this team losing so many people? And the manager — who has been abusive for years — is suddenly a “retention problem.”
“Bas numbers dikhe toh action hua.” Only when the numbers showed up did action happen. This is not a complaint about delayed justice. This is a description of how institutions work. The employees who left — they complained. They told HR. They told their skip-level managers. They told exit interviewers why they were leaving. But none of this triggered action. What triggered action was: the turnover number crossed a threshold. The dashboard turned red. And leadership could no longer pretend the problem did not exist.
The employees who are still on that team know what this means. It means the institution did not care about them being abused. The institution cared about the operational cost of replacing them. The manager is not being held accountable for abuse. The manager is being held accountable for turnover. Which means: if the manager can reduce turnover without reducing abuse (by hiring more resilient people, by increasing compensation to offset the abuse, by making it harder for people to leave), the institution will accept that solution. Because the institution’s problem is not abuse. The institution’s problem is the number.
Linda Wachner. The Turnover That Became Undeniable.
Linda Wachner, former CEO of Warnaco, was infamous for publicly demeaning employees. When she was fired in 2001, the New York Times ran a long story listing “one indignity after another” that she allegedly inflicted. Business partner Calvin Klein said, “She is abusive to our people. Verbally, the language is disgusting.”
Warnaco’s turnover under Wachner’s leadership was severe. Senior executives did not stay. Presidents of divisions rotated rapidly. The company went through multiple CFOs. This turnover was visible. This turnover was operationally expensive. And this turnover — combined with financial troubles — eventually forced Warnaco’s board to act.
But here is what the turnover number did not capture: the employees who stayed and endured Wachner’s abuse because they could not afford to leave. The employees who were humiliated in meetings, who were subjected to late-night phone calls demanding they come to the office early the next morning only to be left sitting in a room for hours, who heard Wachner make ethnic and racial slurs. These employees experienced harm. But their harm did not appear in the turnover metric. Because they stayed.
The turnover metric measures exits. It does not measure endurance. And the people who endure — who stay despite the abuse, who tolerate the humiliation because they need the job — their suffering is invisible to the institution. Because the institution only measures the people who leave.
The Measurement Gap as Institutional Choice.
Institutions could measure psychological harm. Institutions could survey employees about their managers’ behavior. Institutions could track complaints. Institutions could monitor health metrics (sick days, stress leave, therapy requests). Institutions could measure performance degradation under abusive supervision. Institutions could do all of this. Some institutions do some of this.
But most institutions do not. Because measuring psychological harm is complicated. Measuring psychological harm requires asking uncomfortable questions. Measuring psychological harm produces data that implicates the institution. Much easier to measure turnover. Turnover is clean. Turnover is objective. Turnover is a number everyone agrees on. And turnover — unlike psychological harm — does not require the institution to acknowledge that it employs abusive managers.
The measurement gap is not an oversight. The measurement gap is a choice. The institution chooses to measure what is operationally expensive and to ignore what is psychologically harmful. And this choice has consequences. The consequence is: abuse is tolerated until it becomes operationally expensive. Victims are ignored until there are enough victims that the institution cannot staff the role. Complaints are dismissed until turnover forces accountability.
The institution could measure harm directly. The institution could intervene before people leave. The institution could protect employees rather than wait for operational costs to exceed political costs. But the institution does not. Because the institution has decided that psychological harm does not count as damage until it converts into a number that appears on a dashboard. And the only number that consistently appears on dashboards — the only number that forces institutional accountability — is turnover.
What the Turnover Number Reveals. And What It Obscures.
119 assistants in five years. This number reveals that Rudin’s workplace is untenable. This number reveals that people cannot stay. This number reveals operational dysfunction. This number forces the institution to respond.
But this number also obscures. This number does not reveal what Rudin actually did. This number does not reveal how the assistants felt. This number does not reveal the assistants who stayed. This number does not reveal the assistants who lasted less than two weeks and were not counted. This number does not reveal the psychological damage, the panic attacks, the dread, the humiliation, the rage, the shame.
The turnover number is evidence of abuse. But it is incomplete evidence. It measures exits. It does not measure harm. And institutions — by relying on turnover as the primary metric of managerial dysfunction — create conditions where abuse is tolerated until enough people leave that replacement becomes operationally expensive.
The turnover number is the only number that forces accountability. Not because turnover is the most important metric. But because turnover is the only metric institutions consistently measure. And what institutions do not measure, institutions do not see. And what institutions do not see, institutions do not fix. Until the number becomes so large that ignoring it is more expensive than addressing it.
Source material: Chapter 1, The No Asshole Rule — Robert I. Sutton (2007). Scott Rudin’s assistant turnover data (119–250 assistants in five years), Linda Wachner’s executive turnover at Warnaco, and the implicit pattern that abuse is tolerated until it becomes operationally expensive.