Background Music: Shut Your Eyes by Snow Patrol
题记：Since Jack Welch, at General Electric used employee ranking to make decisions about how to fire the lowest ranked employees, more companies have moved to a ranking system. The problem is that it’s mathematically unsupportable, can result in having a less able workforce than when you started, and puts employees in a position where their interests are served if their colleagues FAIL. Learn more about the problems from both economic and psychology perspectives.
From Economic perspective: Why Employee Ranking Systems Lead To Disaster
In our white paper entitled “Performance Management — Why Doesn’t It Work we discuss some reasons why most performance appraisal systems fail to add value to organizations. Despite our work, and the work of more illustrious experts such as Deming & Scholtes, most people believe in the usefulness of performance management. In a way that’s understandable, since it CAN succeed in the hands of an excellent manager, and the importance of performance management has been stressed for decades in much of the management literature.
What is more perplexing is the continued use of ranking methods to evaluate employees. Ranking employees, particularly for determining promotion, and pay, or even for providing developmental feedback simply makes no sense. It is not a neutral process, or just a costly process–it is a recipe for disaster. This month we look at why this is so. (Next month we will take on the use of rating systems).
Rankings In Appraisal
The core element of the use of rankings is that employees are compared to each other, and given some number that supposedly indicates whether they are better than, about the same, or less effective than their colleagues. That ranking is often used to determine who will receive pay raises from a limited pool of money, or for other decision-making processes.
The criteria for ranking can range from specific and objective to totally fuzzy and subjective. For example, it is possible to rank sales staff objectively, in terms of the sales generated in a year, and identify the top salesperson, the next best, down to the bottom based on some reasonably meaningful numbers. Or, one can rank people on a set of fuzzy criteria such as “gets along well with team members”.
The Arguments In Favour
There are only a few arguments to support the use of rankings in any plausible way. The major argument appears to be that ranking employees versus each other creates a situation where competition can be encouraged–the assumption being that if staff compete with each other they will push each other to greater productivity.
The second argument is more administrative. Organizations that rely on merit assessments for decision-making on pay levels and promotions need to decide who will get what. Proponents of ranking systems suggest that rewards for productivity should go to the top performers as defined by comparison with their peers. So a ranking system allows organizations to decide to reward the “top 25%” or the “top 10%”. On the surface this makes some sense. Given a limited pool of rewards, shouldn’t the rewards go to the top performers in the organization? We’ll see.
The Arguments Against
Let’s counter the administrative argument first. We want to reward people for the value they contribute to the organization (however that might be defined). The catch is that a ranking system doesn’t do that. It rewards for being better than one’s peers, and that’s a very different thing. The easiest way to show this is to look at an example. We are going to use a sales example with rankings by total yearly sales, because that’s a best case scenario, since we can measure sales objectively. If ranking systems don’t make sense there when we have good data to guide the rankings, they aren’t going to work with more fuzzy ranking criteria.
Let’s take a small group of five people with sales figures as follows:
Our system calls for rewarding the top 20% (one of the staff) with a significant pay raise, while giving a small “average” reward to the middle 60%, and giving no reward at all for the person at the bottom.
Bob gets a big raise while Ken, Mary and Barb get a little, and Fred receives nothing. Does this make sense? No.
If we look at the figures, we see that we are rewarding Bob for his ability to be one dollar better than Ken. In fact the difference among all of the salespeople is small…and this isn’t surprising since we assume a reasonable job selection process where only the best are hired and retained. So what we are doing here is making important decisions based on almost no differences in production because our “system” specifies that we must reward the top 20% with no room to evaluate the absolute contributions.
Apart from the fairness of this, what effect might it have on the performance of Ken and the others?
But here’s the real kicker. Let’s look at the value that each of these people contribute to the organization. Let’s assume that each of the sales staff draws a base salary of $30,000 a year. When we look at the absolute value of each staff member, we see that NONE of them are adding value. They are costing the company more than they are earning. Under a strict ranking system we would still be obligated to pay that top performer his raise, even though Ken is simply the best of the really lousy!
Ranking systems don’t assess value and contribution, even in a best case scenario.
The other argument put forth is that ranking systems encourage competition, and that is probably true. The error with this argument is that it assumes that competition will lead to increased productivity, and increased success for the larger organization. This is rarely the case. Why?
Quite simply, we tend to get the kinds of behaviour we reward. We can set up a system with good intentions, but unintentionally encourage behaviour and actions we don’t want. Ranking systems (and related reward systems) allow for two ways to “win” extra rewards. The first, and the one we would like to see most is for people to work harder, better and smarter and become more productive. By being more productive they can vault over their lesser performing colleagues to receive additional rewards. The second possibility is to contribute to degrading the performance of those competing for the same reward. An employee can vault into the upper echelons of ranked performance by helping others do worse.
This is certainly NOT what we want.
While it is only the most cut-throat employees who will deliberately attempt to reduce the effectiveness of colleagues, the use of ranking and related rewards does push even “nice” people into doing things damaging to the organization. If you reward based on relative ranks, you encourage:
- hoarding of resources so they are “there when needed”
- with-holding of information
- reduction of team-work and helping others
- and generally self-centred and self-serving actions.
1) While ranking may seem to provide an objective means of evaluating (since it can be used to assign numbers to people), the rankings themselves are only as good as the criteria used for ranking. They can be extremely deceptive, making it appear that there is an objective valid evaluation process going on when, in fact, there isn’t.
2) The value of an employee RELATIVE TO PEERS, is irrelevant to the success of any organization. It matters not a bit whether a person is the best or the worst. What does matter is their absolute contribution to the goals of the organization. Ranking doesn’t improve organizations. It only classifies people and does not reflect the actual value of employees.
3) As a form of feedback ranking is virtually useless. If our goal is to develop people, we need to provide specific concrete feedback. Informing someone that they ranked in the top (or bottom) twenty-five percent on something may send some sort of message, but tells the recipient virtually nothing about how to improve.
4) Ranking can be devastating to the morale and trust of an organization. Because it is difficult to rank objectively, employees will almost always disagree with a ranking that places them anywhere but in the top percent in the organization. Employees often perceive the process as unfair and arbitrary. Research has shown that the large majority of people believe they are above average in job performance. Ranking guarantees disagreement.
5) Finally there is the issue of comparisons. In today’s work world, even people with the same job titles in the same “shop” may be doing very different jobs and contributing in very different ways. How is it possible to compare someone who functions as an informal workplace leader to someone who is technically talented but interpersonally unskilled? Both contribute in their own way. It really is like comparing apples and oranges.
The Disaster Part
If some lunatic was to ask you to create an organization full of dissent, back-biting, resource hoarding, secretiveness, lack of trust, etc, you probably would choose to use a ranking format for performance management. You would also have an organization that wouldn’t know who was contributing to the company in any absolute terms and an organization that would have considerable difficulty providing developmental feedback to staff for the purposes of improving performance.
As a final note, somewhere on this planet there are people who use rankings and swear by them. It may be they aren’t looking in the right place to evaluate the overall effects of such a strategy. In rare cases, it may be that they are in fact building positive outcomes. As with many performance management techniques, however, where you find a manager succeeding with a ranking system, we guarantee you will find a manager who would succeed with scribbling performance appraisals on toilet paper. In other words, in spite of the system!
From Psychology Perspective: Ranking Employees: Why Comparing Workers to Their Peers Can Often Backfire
We live in a world full of benchmarks and rankings. Consumers use them to compare the latest gadgets. Parents and policy makers rely on them to assess schools and other public institutions, and sports fans like them for help in sizing up their favorite teams. But what about when rankings are used at the office for appraising staff performance?
It’s often assumed that employees who are benchmarked against each other work harder, to either hang onto a high ranking or raise a low ranking. However, Iwan Barankay, a management professor at Wharton, calls that assumption into question in a new study titled, “Rankings and Social Tournaments: Evidence from a Field Experiment.”
“Many managers think that giving workers feedback about their performance relative to their peers inspires them to become more competitive — to work harder to catch up, or excel even more. But in fact, the opposite happens,” says Barankay, whose previous research and teaching has focused on personnel and labor economics. “Workers can become complacent and de-motivated. People who rank highly think, ‘I am already number one, so why try harder?’ And people who are far behind can become depressed about their work and give up.”
Barankay’s interest in rankings as a motivational tool intensified during the aftermath of the 2008 financial crisis, which “showed us that offering employees financial incentives based on their performance can have unintended consequences,” he notes, referring to the sky-high bonuses earned on Wall Street in the run-up to the downturn.
“The practical question I wanted to answer is: What should employers do to make their employees work harder when financial incentives [aren’t effective] anymore? It is often thought that people care about their status compared to others — that people derive some happiness or dissatisfaction from knowing they’re better or worse than their reference group,” Barankay states. “Of course, rank should matter if money is at stake. But I looked at rank as its own reward. I wanted to find out whether workers truly want to know how they rank against their peers and … if they knew how they ranked, did it cause them to adjust their effort?”
His study involved 330 employees recruited via Mechanical Turk, Amazon.com’s “crowd-sourcing” platform for work conducted and submitted online. Employers post jobs on the website’s listings section — most of which involve piecemeal, routine work, such as organizing photos, writing or editing text, and basic data entry. Prospective employees scroll through the list and select a task they want to complete.
When workers, also called “turkers,” click on a job, they are led to a web page that presents a set of tasks. After completing the tasks, a worker can decide whether to continue on to the next job. The jobs typically pay $.03 to $.50 per task, and tasks usually take between a few minutes to an hour to complete. Among the companies that use Mechanical Turk are Google, Yahoo and Zappos.com, the online shoe and clothing purveyor.
“It’s a platform that represents the new frontier of work,” Barankay states. “The assignments on offer are things you can’t program [a computer to do]. They’re tasks that require human input, but they’re not worthy of [creating] an entire job. It’s a way for employers to get some back-office work done and for workers who need flexibility to make some extra cash.”
According to Barankay, using Mechanical Turk for a field experiment is attractive for a number of reasons. First, it is a natural environment in which to study human behavior in a way that laboratory settings aren’t able to accommodate. Second, the timeframe is short: Experiments can be completed in a couple of hours, although long-term tests can be conducted if needed. Finally, the demographic profiles of the turkers are generally broader than the conglomeration of workers in most companies or in a group of participants in laboratory experiments. “Most important,” Barankay adds, “is that the platform gives you data from the real world. Nothing is more compelling than data from actual workplace settings, but getting it is usually very hard.”
For his experiment, which lasted two weeks, Barankay advertised jobs on Mechanical Turk that involved analyzing images. Nothing about the jobs appeared to be different from the listings on the site offered by the likes of Google or Yahoo. He paid $.05 for every four tasks completed, irrespective of the quality of the responses. In other words, the turkers made money no matter how well they completed the task.
In the experiment’s first stage, Barankay posted two identical jobs, but one offered feedback on the worker’s accuracy at the end of the assignment, while the other didn’t. Because conventional management wisdom contends that people want to know how they rate, Barankay thought the first job would be more popular. Instead, the job without the feedback attracted more workers — 254, compared with 76 for the job with feedback.
“This was a surprising outcome, but it speaks to the paradigm of revealed preferences,” he notes. “Economists are usually very skeptical about what people say they will do. We focus on what people actually choose to do. Their choices convey information about what they care about. In this case, it seems that people would rather not know how they rank compared to others, even though when we surveyed these workers after the experiment, 74% said they wanted feedback about their rank.”
Coming Back for More
In the second stage of the experiment, Barankay randomly divided workers into two groups — a control group receiving no ranking and a treatment group receiving feedback with a ranking. He then sent an e-mail to all of the workers inviting them to return to do more assignments. The content of all the e-mails was the same, except that individuals in the treatment group found out how they ranked in terms of their answers’ accuracy. The aim was to determine whether giving people feedback affected their desire to do more work, as well as the quantity and quality of their work.
Of the workers in the control group, 66% came back for more work, compared with 42% in the treatment group. The members of the treatment group who returned were also 22% less productive than the control group. This seems to dispel the notion that giving people feedback might encourage high-performing workers to work harder to excel, and inspire low-ranked workers to make more of an effort. “This indicates that when people are great and they know it, they tend to slack off. But when they’re at the bottom, and are told they’re doing terribly, they are de-motivated,” says Barankay.
His research also challenges the idea that rankings could provide poor-performing employees with empirical feedback that will dissuade them from staying in their jobs — at no great loss for the employer. “There has been this sense that people on the bottom will realize they’re in the wrong job and just leave, which would also be beneficial to the company,” Barankay notes. “There is also the hope that giving feedback about rank helps retain the top performers. But that’s not the case. Perhaps this is because top performers move on to new challenges and low performers have no viable options elsewhere.
“Of course, in some instances, providing feedback will be a motivational tool that entices people to work harder. But overall it does not appear that way,” he adds. “So the question becomes: Is [ranking employees] worth it?”
Barankay notes in his paper that future work needs to be done to test the effect of rankings in other work environments and “also to explore whether the underlying parameters can be recovered to pinpoint more detailed mechanisms in the data. Only then can we establish if targeted feedback that takes into account the underlying [differences among workers] can be established to generate a positive casual effect on performance.” At this stage, however, “the aggregate result is that feedback about rank is detrimental to performance,” he writes.
But while his research shows that giving feedback about rank doesn’t necessarily lead to increased productivity, it is well documented that tournaments, where rankings are tied to prizes, bonuses and promotions, do inspire higher productivity and performance. When considering these two things together, a lesson emerges, he notes.
“In workplaces where rankings and relative performance is very transparent, even without the intervention of management … it may be better to attach financial incentives to rankings, as interpersonal comparisons without prizes may lead to lower effort,” Barankay suggests. “In those office environments where people may not be able to assess and compare the performance of others, it may not be useful to just post a ranking without attaching prizes.”
The critical lesson for employers is to consider how each employee will respond to feedback and then decide whether sharing that information will be beneficial for everyone involved. “A good employer knows its employees very well and should have a good idea how they will respond to the prospect of being ranked,” he says. “The key is to devote more time to thinking about whether to give feedback, and how each individual will respond to it. If, as the employer, you think a worker will respond positively to a ranking and feel inspired to work harder, then by all means do it. But it’s imperative to think about it on an individual level.”
人们常常认为，彼此之间进行表现对照的员工会更加努力地工作，以保持自己的高排名，或提升位居低位的名次。然而，沃顿商学院管理学教授伊万·巴兰科（Iwan Barankay）在一篇题为《排名和社会锦标赛：来自现场试验的证据》（“Rankings and Social Tournaments: Evidence from a Field Experiment.”）的新论文中，对这个观点提出了质疑。
“许 多管理人员认为，向员工反馈他的绩效表现与同事对比的排名结果，能激发员工变得更具竞争精神——也就是说，他们会更加努力地工作，以便赶上甚至超越别人。 但是，事实是，结果恰恰相反。”巴兰科谈到，他此前的研究和教学一直专注于人事和劳动力经济学。“有的员工会变得越发自满，而有的员工则会因此而消沉。那 些在排名中名列前茅的人认为：‘我已经是第一名了，干嘛还要更加努力？’而那些排名远远落后的人在工作中则会变得抑郁消沉，并放弃努力。”
“我想要回答的现实问题是：当金钱刺激不再有效时，雇主应该采取什么办法来促使员工更努力地工作呢？大家通常认为，人们很在意自己与他人相比的地位 如何，自己的地位高于对照群体会给他们带来快乐，而地位低于他人则会让人们感到不满足。”巴兰科写道。“当然，如果排名高低还涉及到金钱，那么排名自然就 更重要了。不过，我认为，排名本身就是一种奖赏。我想弄清楚的是，员工是否真想知道自己与同事相比的排名高低……如果他们知道了结果，这是否会促使他们调 整其努力程度呢？”
巴兰科研究了330名通过“土耳其机器人”（Mechanical Turk）招募的雇员的情况，土耳其机器人是亚马逊公司（Amazon.com）的一个“众包”（crowd-sourcing）平台，受雇者在线完成工 作并提交工作成果。雇主会把工作任务发布到网站的工作任务列表一栏，其中多数是零碎的日常工作，比如组织照片、编写或编辑文档，以及输入基础数据等。潜在 雇员可以翻看任务列表，之后从列表中选择自己想要从事的工作。
点击某个工作后，雇员——也称为turker——会进入列出一组工作任务的链接页面。完成这些任务后，雇员 可以自行决定是否继续从事其他工作。完成每项任务通常需耗时几分钟到1小时不等，工作的收入一般是3美分到50美分之间。使用这一众包平台的公司包括谷歌 （Google）、雅虎（Yahoo）以及服装和鞋类在线销售商Zappos.com等企业。
巴兰科认为，用“土耳其机器人”这个平台进行实地实验 很有吸引力，其中有多个原因。首先，在研究人类行为方面，这个平台提供了实验室无法提供的自然环境；其次，本次实验时间较短，两个小时就可完成，不过，如 有必要，也可以进行长期实验；最后，与大多数公司的员工或实验室实验的参与者比较起来，总体而言，该平台受雇者的人口统计学特征更为多样。巴兰科还补充谈 到：“最重要的是，你能从该平台获得来自于现实世界的数据。从真实工作场所得到的数据最令人信服，但是那样的数据通常很难得到。”
在实验第一阶段，巴兰科发布了两种完全相同的工作，但是，其中的一种会在任务完成时将工作结果的准确性反馈给受雇者，而另一种则不会。 因为传统的管理观点认为，人们渴望了解自己会受到什么样的评价，所以，巴兰科以为，第一种工作会更受欢迎。然而，不会给雇员提供反馈的第二种工作却吸引了 更多的受雇者，达到了254人，而为雇员提供反馈的第一种工作申请者则为76人。
“这是个令人意外的结果，不过，这个结果却表明了人们‘显示出来的偏好’（revealed preference）。”巴兰科指出。“经济学家常常感到怀疑的是，人们的言行并不一致。不过，我们的关注点是人们所做出的实际选择。这些选择反映了他 们到底在乎什么。在本次实验中，人们似乎不想知道自己与他人相比的排名情况，虽然在实验结束后我们对他们进行调查时，74%的人都表示他们希望得到有关其 排名的反馈。”
在实验的第二阶段，巴兰科将这些受雇者随机分成两组，一个是不会收到排名反馈的对照组，另一个是会收到排名反馈的实验组。随后，他向所 有受雇者发出电子邮件，邀请他们回来完成更多的任务。发给对照组成员的邮件说明，对他们工作的准确性会进行排名，除此以外，所有邮件的内容都相同。这样做 的目的是想弄清排名反馈是否会对他们完成更多工作的积极性，以及工作的数量和质量构成影响。
在对照组中，66%的人要做更多的工作，而实验组成员则只有42%，并且后者的工作效率比前者 低22%。人们通常认为，排名反馈会激励表现较好的员工更上一层楼，同时，能刺激排名较低的员在工作中工付出更多的努力，可是，这个结果似乎与这样的观点 相左。巴兰科谈到：“这个结果表明，当人们很出色而且也知道这一点时，往往就会松懈；而当排名很低并被告知自己的表现非常糟糕时，人们就会变得消沉、气 馁。”
有观点认为，员工排名会使表现较差的员工形成促使他们离职的主观反应，然而，巴兰科的研究对这个观点提出了挑战。“人们一直有这样一种 认识：排名垫底的员工会意识到自己不适合这份工作，因此他们会主动离职，这种结果对公司也有好处。”巴兰科指出。“还有人认为，排名反馈有助于留住表现出 色的员工。但事实并非如此，原因或许在于，表现出色的员工会继续寻求全新的挑战，而表现不好的员工会因为别无选择而留在公司。”
巴兰科在他的论文中指出，今后还需要在其他工作环境中对排名效果进行测试，此外，“为了准确确定数据中存在的更详细机制，我们还需要研 究排名这一潜在因素是否能得到覆盖。只有到那时候，我们才能确定，考虑了潜在因素（员工间各不相同）的有针对性的反馈，是否能对员工表现产生积极的因果效 果”，然而，在目前阶段，“综合结果显示，排名反馈不利于员工的表现”。
“在工作场合中，如果排名和业绩评价即使在没有管理层干预的情况下也能做到非常透明……那么，根据排名进行金钱奖励就是个更好的方法，因为如果没有 奖励，员工间的比较可能就会导致工作努力程度的减退。”巴兰科指出。“而在无法评估和比较员工表现的工作环境中，只是公布一个没有与奖励措施挂钩的排名结 果可能毫无用处。”
对于雇主来说，至关重要的一个环节是，要考虑每个员工对排名的反应，然后确定发布排名信息是否对涉及到的所有人都有好处。“优秀的雇主非常了解自己 的员工，并对员工将会对排名作何反应很清楚。” 巴兰科指出。“重要的一点是，应该多花时间思考是否应该为员工提供排名反馈，以及每个员工对排名结果会做出什么样的反应。作为雇主，如果你认为某位员工会 对排名做出积极回应，并因此受到激励而更加努力地工作，那么，你当然可以这么做。但是，至关重要的是，你必须根据个人的情况区别对待。”