The war for talent is alive and well. In the finance, technology and consulting industries, the competition for “the best and the brightest” is especially heated on campuses as well as in the halls of competing organizations.
Although stubbornly-high unemployment rates have plagued economists and policy makers, many companies have accurately viewed the problem as qualitative, rather than a quantitative issue. Finding people is easy. Finding the right people is a challenge. Recognizing that the extremes inform the mean, many companies have placed a significant premium on top performers. Mark Zuckerberg drew both praise and criticism when he accurately captured the sentiment in an interview with The New York Times.
“Someone who is exceptional in their role is not just a little better than someone who is pretty good,” he said. “They are 100 times better.” In other words, the difference is not merely a rounding error, it is orders of magnitude different.
“Someone who is exceptional in their role is not just a little better than someone who is pretty good, they are 100 times better.”
If the hiring/acquisition strategy of companies like Facebook is any indication, there is a significant gap in the supply and demand for high-quality resources. Theories abound to explain the scarcity on the supply side of the equation – from education (e.g. lack of science and math skills) to politics (e.g. shortsighted US immigration policy). But what of the demand side of the equation? Are companies somewhat, if not primarily, to blame for their own struggles? In short, absolutely.
A contrarian view of the conventional approach to people exposes the self-inflicted pain felt by many organizations today. The challenge that is faced is one of a fundamental nature, the difficultly in forecasting. More specifically, forecasting performance. In the absence of perfect information (if there is such a thing), organizations are stuck with looking at indicators of performance, akin to a financial analyst looking at financial statements to gauge a company’s worth. In the sense that these indicators attempt to approximate performance, they become proxies for performance. However, as is the case with any proxy, there are good and bad proxies. The more the proxy resembles that which it is supposed to approximate (the “actual), the more effective it is as proxy. It is not to say that the use of proxies is a poor practice in general. The use of proxies may be unavoidable or beneficial. The use of a seller’s rating on eBay may be a more effective means of determining the reliability of a seller than physically inspecting their operations, meeting them in person or sampling every item they have for sale. In the case of the people problem, the challenge is not the use of proxies, it is the use of proxies with little to no explanatory power.
Conventional approaches to people place a premium on tenure and stability rather than skill and expertise
Conventional wisdom places a premium on experience. Not surprisingly, this logic underpins a key aspect of many corporate hiring practices. One can reasonably assume that proficiency improves with practice. There are many theoretical and empirical works to support this notion (e.g. learning by doing, Bruce Henderson’s experience curve, etc.). A doctor who has performed hundreds of successful surgeries, should be more proficient than one who is attempting their first operation. A long track record of quality tasks complete is generally a good indicator/predictor of performance. Many organizations carry this logic a step further, placing a premium on an individual with five years of experience over an individual with one. This is where the logic breaks down. A measurement of task duration (e.g. days, months, years) is poor proxy for performance. For reasons that will be expanded upon below, quality tasks complete is. The duration of a task/job/function/title held is a gauge of, well, time elapsed.
In the lexicon of the typical organization, experience refers to the number of years in a role or function. It is implicitly assumed that additional years equate to additional expertise. This logic falls short on both theoretical and empirical grounds. A taxi cab driver operating in Hickory, NC, may have been in business for the same number of years as a taxi cab driver operating in Manhattan, but really cannot be said to have the same amount of experience. The introduction of expertise as a measure of quality tasks complete helps differentiate the two ideas. Standout high-school basketball player, Lebron James had far less experience but far greater expertise than many professional players. By using tenure-based measurements as the yardstick of experience, organizations have a built-in bias against meritocracy. There is a significant difference between experience (time) and expertise (repetitions). Although it may be expedient to use experience as a proxy for expertise, the practice skews incentives, distorts compensation and promotes mediocrity.
The limitations of experience, as defined by the typical organization, are theoretical and empirical. It is not difficult to find examples of superior performance/proficiency by individuals with less tenure. Although significant outliers (e.g. prodigies) often come to mind, less dramatic examples abound in the halls of every organization. A highly-motivated and hardworking employee can quickly surpass his/her peers in a relatively short amount of time. Experience defined through duration denies both qualitative and quantitative differences in effort. By defining experience as time, all people are forced to progress at the exact same rate. An individual working 40 hours a week, can claim the same “experience” (in a given period, say, one year) as one sacrificing nights and weekends to put in 80 hours a week. As the elapse of time cannot be influenced by working longer, more efficiently or with higher quality, it is the very antithesis of meritocracy. Additionally, there is often an inverse relationship between tenure and performance. Performing the same function for years may be a better indicator of complacency and lack of promotion than deep expertise in a field.
Standard resume and interview processes are a poor predictor of future performance in a job, but remain the most common vetting tool of organizations.
The standard vetting process for potential candidates involves a heavy dependency on paper resumes and behavioral interviews. Although a set of strong references or a referral from a known associate can play a role in the hiring process, many organizations are willing to make a hiring decision based on a one-page document and feedback from an interviewer in a 30-60 minute session. Put in more explicit terms, the majority of organizations view the standard resume/interview process as a valid indicator/predictor of performance. Despite significant research to the contrary, the practice remains the “gold standard” (Austrians economists, please excuse the irony of this term) of organizations everywhere. To be fair, many organizations understand and accept these limitations. The struggle comes in defining and executing a new approach. In this case, the classification of “contrarian” is in the consistency between thought and action, not a philosophical difference in perspective.
Expanding upon concepts introduced in the previous section, the key challenge in the hiring process in the use of proxies to forecast performance. Hiring managers must make a determination about an individual’s likely performance in a function without ever seeing the candidate perform the function. In many ways, it is similar to attempting to pick a winning horse before a race. Gamblers look for indicators (with varying degrees of accuracy, science and superstition) of future track success. In the same manner that organizations use experience as a proxy for expertise, resume/interview performance serves as a proxy for job performance. While the use of proxies is not fundamentally wrong, the further the proxy is from that which it is attempting to approximate, the lower the value of the proxy. Additionally, if significant evidence exists of a lack of correlation (or even inverse correlation) between the proxy and the actual, the case for rejecting the proxy is strong. In the case of conventional views on interview/resumes, both conditions hold.
A strong resume and performance in an interview is just that – a demonstrated ability to perform in an interview and represent oneself on paper. Interviews are more often a reflection of likeability than they are a predictor of performance. Pointing out limitations in the predictive powers of resumes and interviews is no longer overly controversial (see What The Dog Saw by Malcom Gladwell), however apologists of the practice point to the fact that (a) verbal and written communication skills are an important aspect of many jobs, such as client-facing positions and that (b) “likeability” is an important factor in the hiring decision (e.g. teaming, culture, social IQ, etc.). Even if the intended/desired predictive abilities are not perfect, there is value in the resume/interview processes’ ability to gauge these other important aspects.
These arguments fall short for a few reasons. The strongest reason also happens to be the most apparent. The definition of performance is not the same in all jobs. While it is true that a resume/interview process will be a better proxy for some jobs than for others (sales versus say, protection services), this ignores the obvious fact that individuals can and typically do receive support in the form of editing, proofing and coaching on resumes and interview preparation. The hiring manager has no way of knowing if the words on the page are that of the candidate or those of a confederate. Similarly, in the typical 30-60 minute behavioral interview a candidate can easily project a different personality, energy level and motivation. How often is the person in the interview different than the person who shows up on day 60 of a job? There are hundreds of books, videos and resources on mastering interviews and increasing likeability. Even if there is a high correlation between likeability and performance (e.g. sales), a 60-minute timeslot is a ridiculously short amount of time to make that assessment. By continuing to subscribe to a process that provides such powerful incentives to stretch the truth, organizations have put themselves at a tremendous disadvantage in finding the right people.
Significant increases in salary as a means to recruit and acquire external talent distorts incentives and contributes to attrition of top performers.
Once the candidate has been sourced and vetted, an offer of increased compensation is typically a key aspect of “getting to yes.” Once again, the standard practice is problematic. Although significant flaws exist in the typical performance management approach of many organizations, the scope of this critique will be limited to the recruitment/negotiation side of compensation. In regards to the approach to compensation, the root cause is two-fold. First, there is the (same) challenge of proxies in an attempt to forecast value. Second, there is the added and perhaps more important element of negative side effects (externalities, for the econ nerds) of the process to the rest of the organization. As compensation is a critical aspect of employee satisfaction, positive actions taken to attract talent often create negative effects on retention.
The concept of value is notoriously controversial; its definition is hotly debated among economists, sociologist and philosophers. In financial markets, an investor’s approach to assessing value, or worth (if one subscribes to the notion) of a given security can define their entire investment approach. “Animal spirits” aside (which also can be viewed as a challenge of execution, rather than strategy), a different approach to value can cause two successful investors reach opposite conclusions about the value of a given security. Apply notions of value and worth to human beings, and the opinions and controversy grow exponentially.
In the context of the compensation question, there are often two implicit calculations at work. First, what is the person worth (question of intrinsic value)? Second, what will it take in order to entice a person to accept the job? Once again, organizations rely on some fairly dangerous proxies. To accurately answer the first question, a cost/benefit calculation is at work. Organizations attempt to pay less in cost than they receive in benefit from an individual’s effort. That being said, calculating the benefit from one individual can be a challenge, especially in industries in which subjective factors are at work. Sales, gambling and hunting contain fairly objective measures of performance, but many industries and positions are not so black and white. This is to say nothing of the ability to accurately assess value in the future. In order to avoid the intrinsic value question, companies have settled on the classical economist’s definition of worth; it is whatever the market is willing to pay.
Borrowing still more from the field of economics, this analysis can be carried further still. As the market for a candidate’s talent is defined by the current employer, the hiring firm and any other potential employers, assessments of value are generally anchored to current compensation level (what the current firm is willing to pay) as well as insights as to what another (typically competing firm) will offer. In order to gain a commitment, a firm offering a 25 – 50% raise is not unheard of. Although there can be many factors that go into a decision to leave an employer, increased compensation is typically a key consideration. Additionally, factors such as better work-life balance, culture and other non-monetary motivations do not generate the same distortion effects that changes in compensation can create.
As a means to attract and hire candidates, the practice of offering significant salary increases can be effective. Money talks, full stop. What often goes unstated, however, is the direct and negative impacts that such actions can have on morale, perceptions of fairness and retention of existing top performers. Assuming compensation levels are similar among competitors, bringing in a new resource at 125% of his/her prior salary will likely create a significant delta between the new hire and his/her peers. Although firms tend to normalize salaries over time, this significant delta can create resentment in the short term. As more firms in an industry adopt/maintain the same practices, the temptation for a top performer to test what the market will bear will only increase with the influx of higher-paid new hires. In addition, as firms demonstrate a greater willingness to pay a 25% premium for a top performer from another firm than they are to offer a 25% raise to a top performer in their own firm, the incentive to shop around is enormous. If firm A is willing to pay an additional 25% for a top performer from firm B, and firm B is willing to pay an additional 25% for a top performer from firm A, at what point have we reached a zero-sum game? In other words, are firms actually improving their talent pool or just playing a shell game with resources?
The challenge of forecasting is compounded due to the preference for new, potential top performers over existing, proven performers. Not only are companies willing to use poor proxies as indictors/predictors of performance, they are willing to place a premium on them. When the fully-loaded costs of a new resource are considered (e.g. salary increase, executive search, interviews, HR teams, etc.), the costs of recruiting and hiring a new resource can easily exceed 200% of the cost of a current employee. The fact that new employees can be very expensive is only part of the problem. A certain portion of new recruits will not perform as advertised, will voluntarily leave, will be fired, etc. Any benefits from bringing new talent must be measured against the fully-loaded costs of acquisition, including an adjustment for “failures” along the way. As internal resources are known commodities, even an average performer can be reliably counted to be average. The bias of companies for the new and unproven over the tried and true is puzzling.
Where to go from here?
As stated in the introduction of this paper, to point out the limitations of the conventional approach to people is not overly controversial. Many organizations recognize the challenges, but have yet to identify and implement a better approach. The solution to the people problem starts with the recognition of the root cause of the issue, the difficultly of forecasting and the use of proxies. The relation between any proxy (e.g. resume, interview, recommendation, etc.) and that which it attempts to approximate (the “actual”) can be thought of as a spectrum. If the right end of the spectrum consists of the actual function, the forecast power of the proxy diminishes as one moves farther to the left. At the leftmost point, there is zero correlation between the actual and the proxy (e.g. behavioral interview examples discussed earlier). The closer to the actual, the more powerful the proxy.
With this mental model in mind, the solution to the people problem consists of a few simple rules:
First, a proxy with no correlation (or inverse correlation) to the actual must be rejected.
It is on the basis of this rule that the conventional interview/resume processes should be thrown out for most organizations. As something such as soft drink preference has little to no correlation with performance on the basketball court, it would be foolish for a basketball coach to spend time on collecting that information. Statistics such as vertical leap, height and wingspan are much more appropriate proxies in forecasting basketball ability.
A more nuanced view on this point is that certain proxies may lose their forecast ability beyond or below a certain point. There may be a correlation between GPA and job performance in general, but the correlation may become weak past a certain point. Once you get above a certain GPA, other factors become important points of difference. Additional data points certainly help, but there are many intangibles that are not easily translated into metrics. Certainly factors such as intrinsic motivation, comfort with uncertainty, genuine curiosity, risk tolerance, aptitude and emotional intelligence are just a few such factors. The behavioral interview attempts to tease out these notions but ignores the obvious fact that people can and do misrepresent their abilities in these areas. Additionally, the more factors one adds to the hiring equation, the more complex any structured approach becomes.
One of the most radical and effective ways to deal with the challenges above is to apply the “wisdom of crowds” concept to the problem of people. At the center of this theory, popularized by James Surowiecki (see The Wisdom of Crowds) is that large groups of people have better predictive capabilities than individuals. More specifically, large groups of amateurs consistently show better predictive power than an individual expert. In his annual speech to Columbia MBA students, Warren Buffet has been known to present the students with a simple mental exercise. If you could invest in the lifetime earnings of one of your classmates, who would it be? He goes on to explain that it likely would not be the one with the highest GPA, the one with the highest-paying job, the most well-liked, or the one with the most impressive resume. It would be an individual with a collection of skills, attributes and principles that, observed over the span of two years, would give a good indication as to the individual’s capabilities. Although there might be some debate about the top spot, there would be a high degree of consistency in the selection of the top five. While Buffet’s point is more about the importance of principles and ethics, there is another important factor at work here. The group of peers would likely beat any hiring process/organization in their predictive capabilities to identify the most valuable student. How often do we see the high performing individual who falls through the cracks of the HR process, but is really a hidden gem? Rather than relying on the opinions of the professionals (HR), perhaps it is time to rely on the group of amateurs, classmates and peers.
Second, more powerful proxies (further to the right of the spectrum), should be favored over less powerful ones.
In our basketball example above, the use of soft drink preference was rejected in favor of statistical metrics such as height, vertical leap, etc. They are more valid proxies as they are more closely correlated with basketball skill. Aside from statistical proxies, proxies that simulate reality can be a more effective predictor/indicator of performance. Forgive the continued basketball analogy, but watching a player perform skills or athletic drills is a better approximation of game performance than the statistics described above. Many companies have moved towards case studies, projects and technical interviews as they attempt to simulate on the job performance. These simulations are a better proxy than behavioral interviews alone.
Finally, whenever possible, the use of proxies should be substituted with actual performance.
Simply put, the easiest way to predict performance in a job is to observe the individual do the job. Asking a candidate for a demolition crew to explain his hammer-swinging prowess on paper leaves much to the imagination. While watching him play a game of whack-a-mole may be a little better, there is no substitute for simply handing him a sledge hammer and placing him in front of a block of granite. Simulations are great, but even the best simulations are not the real thing. The distinction between game players and practice players is widely acknowledged in sports, drawing a difference between those who may practice poorly, but show up on game day (and vice versa).
The same concept applies to our people problem. Recruiting, paying and hiring based on performance in a simulation is limiting, no matter how sophisticated the simulation. Coupled with the fact that organizations are very hesitant to fire people, the problem is compounded. Internships and referrals are a helpful step in the right direction, but adopting a universal temporary-to-permanent model or a probationary period goes even further. In addition to the ability to observe and assess individual skills, interaction with other people and the organization can also be observed. This is a notion that is significantly lacking in virtually all hiring simulations.
In addition to enabling observation in the real world, temporary-to-permanent models and probationary periods have the positive effect of increasing the assessment duration of a candidate. It is difficult to determine the better team based on one play or even an entire game. Watching the team for an entire season will yield better results. In the same manner, performance assessments based on observation over a period of days are not as reliable as those that have been based on a period of weeks or months. It is not to suggest that a long probationary period is necessary in all, or even any, cases, but it is important to remember that performance may vary based on specific circumstance. Increasing the assessment duration increases the likelihood of a variety of circumstance.
A reasonable critique of the above proposal may include the potential for prospective employees to reject changes to conventional hiring practices. If firm A is offering a guaranteed position after an interview and firm B is offering a possibility of a full time position through a temporary-to-permanent model, it stands to reason that firm B will be at a disadvantage in recruiting the best people. This is an oversimplification and ignores the multitude of factors that go into the decision to accept a job. Getting the hiring decision correct with greater consistency will create a better working environment and culture, a key component of employee satisfaction. This is to say nothing of the efficiency gains from better people, less waste on poor performers, etc., which has a direct impact on the ability of organizations to pay top dollar. While the prospect of letting people go can be scary, it is important to note that industries and firms vary significantly in this regard. Advertising firms tend to ramp up and down with the addition/subtraction of accounts. It is not viewed as a universally bad thing to have been let go at some point in your career. It is expected. Additionally, there is nothing wrong with not quite making the cut at an elite firm, particularly one that develops a reputation for excellence in recruitment and people management. Top performers are willing to deal with the possibility of falling short, and many firms have applied these principles successfully. Under Jack Welch, General Electric (GE) developed a very solid reputation, while maintaining a policy of letting go of 10% of their workforce (bottom performers as defined by the company) each year. Many investment banks only offer a select group of analysts (entry-level) the option to proceed directly to the associate level after their third (elective) year, effectively pushing the majority out to other firms or business school. The ability to recruit and retain talent is not negatively impacted in these cases.