How to predict attrition before it even happens
A short primer on the use of survival models in HR.
Most of us use retention metrics to understand the flow of talent in our organizations.
It’s a good metric — it tells you a lot.
It tells you the number of people leaving. Whether they choose to leave. Whether we care about them leaving. And even, which area of the business are they leaving from or from who.
What it doesn’t tell us is who will leave next.
The retention metric is a lagging indicator of attrition. It tells us how many people have left the organization already. It does not tell you too much about who is next in line.
This is where advanced people analytics come in.
And no, I am not talking about an ARIMA forecasting model. Using ARIMA on attrition data is like putting a finger in the air.
I am talking about survival analysis.
“What is survival analysis?” you ask.
Survival analysis is an analytic technique estimating the likelihood of survival over time. This analysis uses the information about you today to determine how likely you are to live or die.
Dark, eh?
Survival analysis comes from medical research. Medical researchers tried to figure out how likely a patient is to survive cancer or heart disease. Hence, a more gruesome name for the model.
Its main premise of survival models is:
every day you live increases the probability you will die.
Or, at work, for every day you stay with the organization, your probability of leaving goes up.
Hence why this model is perfect for estimating three big things in retention:
- What is the average predicted rate of retention in your organization?
- What is the predicted likelihood of retention for each individual?
- What can improve your retention rate?
Ah, now we are talking.
You see… Survival models look at your organization to predict attrition. You can then use this information to design strategic interventions focused on retention. You are helping people to make the right decision — to stay with you!
And, all you need are two things for everyone in your organization:
- Tenure: How long have they been here?
- Retention: Are they still here?
This is the most basic form of the model. It allows you to build a simple survival curve to the exact probability of survival based on time.
The model becomes more complex when you introduce other information about each individual.
You can include gender, performance, compensation, and engagement for each individual into the model to see which factors predict people’s retention.
Let’s go through the three outcomes of the survival model!
What is the average predicted rate of retention in your organization?
With the simplest model, you can learn the probability of staying at different tenure. For example, in the figure above, at a 1-year mark, you can expect that about 60% of your workforce will still be around. If things do not change, you need to plan for 40% attrition, which is horrendous.
You can go more granular and test your departments to see where people are leaving. This will allow you to intervene immediately to prevent extra attrition.
You can go even more granular.
What is the predicted likelihood of retention for each individual?
Yes, you guessed it; the probability is the composite of individual probabilities. Hence, each individual in your organization has a probability of survival. This probability is a result of their tenure and other factors (e.g., engagement).
Now, if you know your top employee has only a 70% probability of retention, what will you do?
At the very least, you will sit down with them, probe their intentions, and see what you can do to help them decide to stay.
This is where survival analysis becomes especially useful.
It allows you to intervene at the individual level and improve retention.
But, what can improve your retention rate?
Survival models make life even more interesting when they reveal systemic issues.
Predictive modelling can show women leave faster than men. Intervene to make sure women are staying. For example, consider introducing greater sponsorship for women if this is a reason they make their decision to leave.
The model can show that people are moving on for new career opportunities. Create career paths to help people see where they can go.
Employees with a high number of stock options are staying. It sounds like stock options are a strong strategy — keep that up.
What you can find is amazing!
Assuming you did not put garbage into the model, of course.
But more, what you do is key.
Let’s be honest; retention never improved from thinking about retention.
It improved from the action.
Hence, think about how you will use survival models in your people analytics journey.
These models, in turn, will help you survive.
