What causes employee turnover
As much as 80% of employee turnover results from bad hires, and annual employee turnover in the US exceeds 50 million people. That means that approximately 40 million people leave their jobs – voluntarily or otherwise – because of bad hiring. Wow!
The cost of a bad hire depends on the role and many company-specific factors, but it’s commonly estimated to be around 2.5x times salary and can therefore easily amount to hundreds of thousands of dollars. Again, Wow!
Now for some good news. We know what causes the majority of employee turnover – bad hires – and we know what causes bad hires – hiring practices that don’t predict performance, namely résumés and unstructured interviews.
How to predict employee performance
In 1998 psychology professor Frank L. Schmidt from the University of Iowa published a paper titled The Validity and Utility of Selection Methods in Personnel Psychology. It summarized 85 years of research into personnel selection and compared 19 selection methods for predicting job performance. The research concluded that the single best way to predict job performance was to combine GMA and with samples. GMA stands for general mental ability, and a GMA test is what we commonly refer to today as a cognitive assessment. Work samples are described as follows:
“Work sample tests are hands-on simulations of part or all of the job that must be performed by applicants. For example, as part of a work sample test, an applicant might be required to repair a series of defective electric motors.”
In other words, work sample are skills assessments. Back in 1998 professor Schmidt didn’t envisage that work samples could be delivered online in a automated way, but the concept is exactly the same: see how someone can do the job, literally.
The only way to really know whether someone will be great at their job is to evaluate their actual performance after they’ve done the job for a while. Given recruiters don’t have that luxury – recruitment is about working out how could be great at the job they’ve applied for – they have to do their best to reduce uncertainty as much as possible. Schmidt’s paper boils that down to two selection methods used together – a cognitive assessment and a skills assessments. In other words, the best way to predict performance is to hire smart people who can show – literally – they can do job-related tasks. Who would’ve thought that hiring could be so simple? If you want to know whether someone can design, ask them to design something relevant. And make sure they can think.
Selection methods that contribute to employee turnover
Personality traits were found to not be strong indicators of performance with the exception of conscientiousness, which ranked very highly as a predictive trait. Reference checks, years of job experience and assessment centers were all proven to be poor predictors. Years of education was found to be hardly predictive at all.
85 years of research prove that résumés – summaries of a person’s work experience and education – are entirely ineffective at predicting job performance. Structured interviews were found to be quite predictive, especially when combined with cognitive assessments. Conversely, unstructured interviews were considered far less predictive. It is therefore no wonder that hiring decisions made based on résumés and unstructured interviews result in bad hiring decisions and, consequently, high employee turnover.
Job auditions are the new work samples
Companies are increasingly turning to candidate selection methods that show them how candidates actually perform. Skills assessments that test on-the-job skills are often referred to as job auditions, and they attempt to put candidates in situations they’d normally face on the job.
We have seen living proof that companies utlizing skills assessments end up hiring people who perform at a higher level, in addition to the huge efficiency gains they benefit from. It’s not easy to abandon a familiar process and adopt something new. It can be scary. But years of research show exactly which selection methods predict performance, and technology now enables them to be deployed a scale.