I totally agree with this Jackson and I love this callout and push for more causal inference approaches. I'm trying to push in this direction in my own work and it has been a mixed bag. So many HR professionals and business leaders (my stakeholders) are used to traditional metrics such as turnover sliced and diced by variables such as tenure, division, job code, gender, etc. I think a key first step is to start to consider other variables that could have a causal relationship which are not the traditional variables that are easy to get because they are more often standard data fields in HR systems. Are you aware of specific case studies or examples in people analytics that demonstrate some of the methods you describe? I think that it would be crucial for the field to show what is possible and get people out of the traditional metrics and analyses.
Thanks Willis, great questions you outline here. The tricky part about metrics (other than the methods themselves) is that spotting opportunities to use them is more art than science, so to your point examples may be key.
One I have from a prior job is taking an advantage of a natural experiment where an office had closed down due to (purely) financial considerations- leaving many sales and service employees to become remote workers basically overnight.
We were able to do a difference-in-differences comparing turnover to other people with the same jobs in other offices to evaluate the impact of remote work on turnover and performance. This allowed us to confidently shift roles to remote early in the pandemic where other companies where flailing, trying to use personal opinion or conventional wisdom to figure out what works.
I totally agree with this Jackson and I love this callout and push for more causal inference approaches. I'm trying to push in this direction in my own work and it has been a mixed bag. So many HR professionals and business leaders (my stakeholders) are used to traditional metrics such as turnover sliced and diced by variables such as tenure, division, job code, gender, etc. I think a key first step is to start to consider other variables that could have a causal relationship which are not the traditional variables that are easy to get because they are more often standard data fields in HR systems. Are you aware of specific case studies or examples in people analytics that demonstrate some of the methods you describe? I think that it would be crucial for the field to show what is possible and get people out of the traditional metrics and analyses.
Thanks Willis, great questions you outline here. The tricky part about metrics (other than the methods themselves) is that spotting opportunities to use them is more art than science, so to your point examples may be key.
One I have from a prior job is taking an advantage of a natural experiment where an office had closed down due to (purely) financial considerations- leaving many sales and service employees to become remote workers basically overnight.
We were able to do a difference-in-differences comparing turnover to other people with the same jobs in other offices to evaluate the impact of remote work on turnover and performance. This allowed us to confidently shift roles to remote early in the pandemic where other companies where flailing, trying to use personal opinion or conventional wisdom to figure out what works.