In Part One, I outlined how an “average” standard deviation in worker productivity can amount to huge dollar costs for a company. In many cases the amount is large enough to make a difference between organization profitability and bankruptcy. I also said that if you don’t see or can’t measure productivity differences, then there is no reason to change your hiring practices. In this article, I’ll continue summarizing some of the hiring research of Schmidt and Hunter* as it applies to the cost of a bad hire. The Idea of Individual Differences Basic estimates of differences in individual productivity range from as low as 19% to as high as 48% of the average salary for the job. For example, an IT professional or manager paid $70,000 per year who is in the top producing 16% would produce about $33,600 more than average. Someone in the bottom 16% would produce about $33,600 less than average. Again, these are average differences in performance. They could be much greater. The percentages represent “standard deviations” of performance. Rank-ordering employees, calculating the average productivity, and then setting a cut-off point at the bottom 16% and top 84%, respectively, determines the size of a standard deviation. The area from 16% to 50% and 51% to 84% is a “standard deviation”. It is where 2/3 of the people fall. The goal of a recruiter should be to continually work to narrow the standard deviation (i.e., reduce differences in performance) and raise the average (i.e., increase productivity). Narrowing the Differences and Improving the Average There are both internal factors and external factors that affect the standard deviation and average productivity. The major ones are:
- Size of the available labor market
- Organizational “attractiveness” (i.e., pay, benefits, culture, location, etc.)
- The organization’s ability to measure performance differences
- The financial impact of these performance differences (e.g., type of job or the average number of years workers remain on the job)
- The number and skill of internal personnel who manage the hiring process
- The accuracy and completeness of the job model
- The ability of each hiring tool to accurately predict performance
Working From a System Accuracy starts with looking at each job as a “whole system.” This means working from a model that considers the whole job and the whole person. Again, when performance in all jobs is statistically analyzed, it tends to fall into four different areas. These are:
- Cognitive ability: The ability to think, problem solve, learn, technical knowledge, etc.
- Planning ability: The ability to plan, organize, sequence, etc.
- Interpersonal ability: The ability to get things done through others, coach, sell, resolve problems, etc.
- Attitudes, interests and motivations: Likes and dislikes associated with the job
Every job needs certain skill levels in each area. If a recruiter skips any area, skills in that area are left to “chance.” For example, technical managers tend to place heavy emphasis on technical skills (i.e., a cognitive ability) but readily admit that technical people fail more often because of poor interpersonal skills, an area they seldom measure. Call center managers, on the other hand, tend to skip attitudes, interests and motivations, even though customer service representatives tend to quit because they do not like the work. Skip it, or fail to accurately measure it, and you will get performance differences. Setting the Level It is important to remember the “Goldilocks” principle: Some applicant requirements are too high for the job, some are too low, and others are just right. Setting your hiring target too low may lead to performance problems, while setting it too high might lead to early turnover and dissatisfaction. I have known many organizations that recruited from the best schools and hired folks with the highest grades. What did they get? People who started off like a rocket, but became discouraged and quit when they weren’t promoted to president within two years. This is an example of setting an applicant goal too high. Other firms set their targets too low. What did they get? A ready supply of people with low productivity. Setting just the right level can be illustrated by an example. Most of us would think that smarter people tend to be better performers. But that is only partially true. Intelligence and performance do “track” together — but only to a point. When people have more intelligence than the job requires, performance becomes unstable. The challenge to each hiring organization is to find the “breaking point” where performance peaks. There are performance “breaking points” for each of the four major performance groups. Accuracy Throw away silly tests that were not designed expressly to predict performance. I once did a study that showed a simple test did a great job of predicting performance in one sales job, but failed miserably in another. Why? The factors measured by the test and the factors required for job success were entirely different from job to job. This is why each organization needs to validate each test for each job. Using someone else’s validation results seldom works unless their job is almost exactly like yours. A good validation study is important, because it tells you that high scores on the test are associated with high performance, and low scores are associated with low performance. But these types of studies are seldom done. What kinds of studies represent bad hiring practices?
- Studies that use small groups. They are NOT valid because they don’t statistically represent scores of a large group.
- Comparing applicant scores with a “average scores” of special groups such as criminals, engineers, sales people, accountants, etc. They are NOT valid because, for example, they do not predict how a person applying for a job in engineering will perform as an engineer.
- Taking validation results from jobs that are significantly different from your jobs. They are NOT valid because the jobs are very different.
Best practices mandate that you conduct studies that show your test scores predict performance for your job. High scores equals high performance. Low scores equals low performance. Who cares if a personal profile “matches” a large group of people with mixed skills? Don’t you really need to know whether the applicant will be a high or low producer? Thoroughness One test is not good enough. You need a validated tool to accurately measure each job area. These tools include: realistic job previews, learning and problems solving tests, planning tests, one-on-one simulations, job samples, case studies, behavioral interview techniques, situational interview techniques, and tests of attitudes, interests and motivations. You can pick and choose among these tools depending on the thoroughness and accuracy you need. A clerical position, for example, may not require more than one or two tools. A manager or professional position may require four or five. In the old days, administering tests could take several days. Modern tools show that we can determine job skills within a few hours. Conclusion Hiring costs are more than paying recruiting fees, placing advertisements or turnover. They are deeply imbedded in variable levels of productivity. A recruiter is the talent scout for the organization. If he or she uses inaccurate tools or incomplete specifications, then the organization will suffer as a result. Suffer a little? No, suffer to the tune of millions of dollars. * The validity and utility of selection methods in personnel psychology: Practical and theoretical implications of 85 years of research findings. Psychological Bulletin, Sept 1998, Vol. 124, No. 2, pp 262-274.