Or what we can learn about turnover with People Analytics
We’ve been using the word talent for years. It’s very cool to put this word in every speech, whether it’s a fairy tale or not. We hear so much talk about talent management that it seems as if we had witnessed the advent of a new kind of hominid, the homo talentum… which comes to design the future with an irremediable modernity and style that no one else has.
And evidently neither all millennials have talent, nor does the twenty-something talented stop being a fairly predictable homo sapiens in terms of their motivations and behaviors. Of course, this must be understood in the context in which it develops, and it is clear that the technological environment in which these young people have grown up marks a different management strategy. But enough is enough, let’s not exaggerate, the genetics of the twenty has not mutated by the existence of the iPad, the VUCA environments, or the new business models of the digital era. The homo talentum has the same evolutionary heritage as you and me, and as the vanguard of its promotion they bring the new paradigms of their generation, but with the same insecurities and biases that we all had when we were 20.
From there we understand the problem of talent retention as a consequence of the moment of the economic cycle in which we find ourselves, in which for some sectors the supply of the labour market is lower than the demand. This is aggravated, however, by the speed with which technological advances take place, generating new professions for which we will not be able to find people with the speed required.
In any case, whether it is a new and dramatic problem, or the logical consequence of the 3rd technological revolution that humanity is experiencing, what is certain is that the only efficient way to approach the retention of people is by analysing the data we have on them.
And this is where the People Analytics perspective adds value. It proposes an empty analysis of opinions, of hasty interpretations and far from the crickety seed of intuition. We must base our decisions on studying the data under the probabilistic prism provided by statistics. According to this, let’s ask the data why they are leaving? who is leaving the company? at what moment?
A first analysis could be then, on the data obtained when we ask the reason why they leave the company. Unfortunately here the cause-effect correlations are not very high (no greater than 0.25), which gives us to understand that there are many and varied reasons why people leave an organization.
However, we can draw some interesting conclusions. On average, if we don’t like our manager, we are more likely to leave. There will be many people who hate their supervisors who are going to stay, and there will also be many people who really like their supervisor and yet will leave. What is interesting is not the individual correlation of this factor (which is not very high) but its comparison with the rest of factors measured, such as salary. We tend to think that people stay in their jobs for money, and my intuition (always misleading) tells me that this is true, but it is surprising to see that it has a weak effect on this type of survey.
Other ways of approaching the problem is to analyze data obtained from the study on how people make the decision to leave the company. This decision can come because we compare the existing work with an alternative that is better, or because we have a personal plan to move when we reach a certain age or experience. In the first case, it could be that another company contacts us or that we have made the decision to look for a new job. This last option may come from being dissatisfied with our current job or because we are aware of more interesting alternatives. Once again we see that there are many different reasons why a person makes the decision to leave, and that are subject in turn to different very particular considerations.
It is also convenient to analyze the rotation data thinking that it is produced as a normal process of searching for the job that fits us and makes us happier (or less unhappy).
In this case we realize that there are a series of implications to take into account and that we will see represented in the data as probabilistic trends. People are more likely to leave work early, and this coincides with their high turnover rates. Conversely, as workers age in their jobs, their turnover rates are lower.
Thus, saving the effect of economic cycles, the data argue that in reality millennials do not move from jobs much more than people of previous generations when they were that age. At age 20 the probability of moving is always greater. This knowledge is useful to us to interpret correctly the rotation ratios without ignoring habitual tendencies that occur in certain collectives.
The scientific analysis of data allows us to make decisions in hiring or other operations in the company from which emanate statistical correlations with rotation.
It will allow us to know if we are leaving high-performing employees, to know if we are hiring them with a certain probability that they will leave the company, or to find out if we could touch something in their environment to diminish the factors that push them to leave. Therefore, the really interesting thing about the application of People Analytics in this area is to try to predict which factors have an adverse impact on job rotation in order to avoid them.
And in this sense, the most basic thing we can do is to compare the percentages of turnover we have over time, in the different units of our organization and depending on the different supervisors. And in this comparison detect which differences are really significant and which are not (test of statistical significance). The problem here is that these dropout percentages do not take into account the tendency that people are more likely to leave early in their stay in the company.
We will therefore have better quality information if we look for statistically significant differences starting from percentages of turnover by cohorts of months (proportion of people that last 3 months, 6 months, a year, etc.).
A better analysis would be to implement on each time interval a statistical technique such as multivariable regression, which would allow us in probabilistic terms to predict who would leave in each period, and which factors have the greatest adverse impact on retention: age? training? distance to home? etc…
But the most professional way to do the analysis is to use the so-called survival models. They are very precise statistical models, similar to those used in medicine to predict, in terms of probability, the years of life in certain types of fatal diseases. It is a question of drawing a curve that describes the rate of permanence in the company as a function of time. Then represent that same curve but segmenting by a certain type of factor potentially correlated with turnover (for example, experience at the time of hiring). We can compare survival curves for different groups of people who have differences in that factor, and thus check whether the graph has risen (lower turnover) or fallen (higher turnover). We can continue to test this analysis with more factors, such as the type of department the employees belong to, the type of manager who directs them, or personal characteristics, etc. and in this way analyze which of them improves the permanence curve in terms of avoiding turnover. The good thing about this model is that we observe all time intervals at once in a continuum.
Talent turnover has a major economic impact on organisations and we have always been tackling this problem based on instinct and hunches. Changes are an intrinsic part of life, changes are happening faster and faster, but we have more data than ever about what people do and we have solid analytical tools.
Let’s ask the right questions and bring the data to those questions. We can bring much more rigour to the process and improve the quality of the decisions we make, rather than content ourselves with explanations of reality that are of little use to action.
The homo talentum has not yet been born.