Attrition versus Retention: Which Should I Use?

Weighing Pros and Cons of Two Important People Analytics Metrics

Ben Teusch recently published a great guide on how to calculate attrition metrics. If you are new to people analytics, it is a great guide, no matter how incomplete it is. Ben describes differences in different types of attrition, such as annualized attrition versus year-to-date (YTD) attrition versus trailing attrition rates. If you are not familiar with these terms, you can stop here and read up on this and then come back here.

An incomplete starter's guide to attrition metrics

Ready to read on?

I find the guide valuable but I do have one beef with it. Ben doesn’t distinguish between retention and attrition, choosing instead to use terms interchangeably and define everything in terms of attrition.

I believe this approach to be less than helpful. When appropriate it is nice to simplify things with interchangeable terms. However, in this case, I prefer to be more precise about the differences in attrition and use their differences to highlight nuances. Some terms, like attrition and turnover, can be used interchangeably because their differences are so small. However, when it comes to attrition and retention there are nuances worth highlighting. How so? Let’s start by defining them.

Attrition is a far more common term and more commonly used. Like Ben, I’m partial to the 12-month trailing rate, which is essentially a 12-month rolling average of a monthly attrition rate. This is calculated by counting all the people who left in the past 12 months and dividing by the average headcount during that period. In what follows, when I say attrition I mean this 12-month trailing rate, which is the most common way I calculate and use attrition.

Retention is not an uncommon term and is often used loosely as a desirable end goal for organizations. How often do you hear executives say “We’ve got to do better to retain our top talent” or “We have a retention problem”? However, a retention rate is less commonly used. To calculate a retention rate you count all the people who have stayed divided by the number of people you started with. A retention rate requires you to define the corresponding time, such as a 1-year retention or 5-year retention rate. Note that this definition of retention rate is essentially the inverse of the cohort-style attrition rate defined by Ben in his guide.

The biggest difference between the two is the groups of people are counted. For the attrition rate, you are counting the people who left the organization at the same time while with the retention rate, you are counting the people who started at the organization at the same time.

Contrary to what many think, a retention rate and an attrition rate are not inverses of each other. So for example, if you have an attrition rate of 10%, you can’t say that your retention rate is 90%. While there may be a loose correlation between the two, they are calculated as two different formulas. Again, because you are talking about different groups of people at different times, you can’t calculate one from the other. That difference is the reason why I prefer not to use retention and attrition as interchangeable terms.

So when would you use one versus the other?

For starters, retention has a positive connotation focusing on how to keep people engaged while attrition has a negative connotation focusing on how to not lose people. Retention is focused on doing things to support your people so they don’t even consider leaving the organization while a study of attrition focuses on what you did wrong that caused people to leave. Retention rate as a metric generally gives you a more proactive focus than the reactive focus associated with an attrition rate. And like many reactive approaches, it is generally more costly to try to fix issues after people have already left your organization than to fix them before they leave.

Second, attrition rate is likely easier to build as a metric than retention rate. The attrition rate is calculated based on who you have right now and who is currently leaving. A retention rate is calculated based on who had at some starting point in the past. You generally need more discipline in your record retention processes and data governance to study historical groups of people based on the time they started.

The attrition rate is a great metric for organizations that are just getting started in people analytics. When your people analytics capabilities are less mature, you are still likely to have the necessary data for attrition calculations in your HR systems. While the same systems can be used for retention rate, you generally need more historical views of data. If you don’t retain data on departed employees, you won’t be able to calculate a retention rate because the data on those who leave is just as important as the data on those who stay. Similarly, retention rate calculations require more work to define the groups of interest for the business problem. How you group people can have a drastic effect on the calculated retention rate.

Finally, because attrition is more commonly used, I’ve found it easier to find corresponding benchmarks from comparable industries and organizations. I’ve found it more difficult to find retention benchmarks. When looking for retention benchmarks, not only would you be looking to find comparable industries and organizations, but you also have to find comparable periods. You cannot benchmark your 6-month retention rate with a 2-year retention rate from the industry. While I do believe benchmarking to be overrated, this is something to consider if that is important to you.

Both metrics can be valuable despite their differences. I’ve summarized these differences in this table for reference.

Regardless of which metric you choose, you can usually calculate them without too much trouble with software tools and display them in dashboards. If do choose to use them as metrics, be sure that you are trending them over time, that they link to your business objectives, and that you are using them to drive the right actions.

And as always, when working with any analytics problem, you do not start with the data. You have to make do with the data you have or work to get the data you need. Data availability should not be your primary driver for choosing metrics. Start with the business problem, then define the relevant questions, and then work back to the relevant data that is needed to answer the question.