Average tenure: I don't think that word means what you think it means

I’ve been meaning to write about “average tenure” as a company metric for a while. It would be a fine metric if it was actually used to measure average tenure (“wow, so-and-so’s team is not very tenured — they may need more supervision than normal as they ramp up), but I keep seeing it used as a proxy for “how long do people stay before they leave?”

I see infographics like the one above that use average tenure to determine “Top Companies for Employee Retention”, and even on the excellent podcast Acquired, they equated an attrition rate to average tenure:

4.5% of their entire workforce every year leaves of their own desire. That translates to a longer than 20-year average tenure.

You get that number by dividing 100%/4.5% = 22.2 years, but that’s actually how long it would take you to lose your entire company at 4.5% / year, if you didn’t hire anyone else.

That’s the real issue with the metric being used to say something about retention — it’s incredibly influenced by how many people you’re hiring each year, since everyone you hire has a tenure of 0. Generally, average tenure says more about hiring patterns than retention behavior.

A classic example of this is Kodak, whose average tenure has been reported to be around 10 years. Google’s average tenure appears to be closer to 2 years by what I found. Does that mean Kodak’s better at retaining people than Google? Not necessarily. Kodak hasn’t been hiring as much in recent years, so an average employee will have been there longer than at Google, who hires more aggressively.

And that’s not even taking into account the fact that if all your attrition is among your new hires, that boosts your average tenure, even with a potentially high overall attrition rate.

So, what should you use instead if you want to look at tenure in the context of how long people actually stay?

Survival curves / cohort analysis: Track what % of people hired in a given year are still there after 1, 2, 3+ years. Once 50% of a cohort has left, you can share that as a single data point (50% of our customer service reps have left within X years of joining).

This is an ideal way of looking at things, but does require a fair amount of historical data and statistical know-how.

Median tenure at time of exit is a simpler approach: Look only at people who left in the past year (or whatever period you like) and calculate the median tenure. Even better, look at 25th, 50th (median), and 75th percentiles to get a fuller picture of the distribution. This approach doesn’t directly account for how long people stay, but across the quantiles you should have a decent picture of when in their tenure people tend to leave.

I’m sure this isn’t the only misused metric out there, so stay on your toes!