There’s a famous story wherein a statistician tried crossing a river based on the knowledge that the average depth of the river was 6ft. If he were alive today, he’d tell you why averages can be dangerous.

However, instead of recounting what happened to this statistician, let me narrate stories from closer to home. Based on an annual engagement survey, the average inclusion score of a team was rated 4+ out of a possible maximum of 5; yet concerns from new hires kept pouring in, and these new hires clearly felt far less included than 4+/5. It wasn’t until both metrics, the average and the individual, were put side by side that we realized that those who’d been around since before the pandemic bonded well together and were extremely good at staying connected. On the other hand, this team had a much smaller number of new hires who unintentionally were left out of most virtual coffee chats, informal huddles and ultimately felt highly neglected despite best intentions. In this case, the average score greatly overrepresented those who felt included, and underrepresented those who did not.

We know averages give us very limited information. Yet, our workplaces and dashboards are riddled with it. An average occupancy rate of 22% for example, tells us very little about attendance patterns and even less about the floors and buildings that could be possibly shut down. It gets even harder when we use averages as comparators across different populations.

For our roles, averages pose an even greater danger. One can’t look at totality and pull out an obvious answer to say this is how everyone is feeling. While the number may be interpreted as things going well or down the drain, everyone at the workplace is going through very different experiences. Trying to condense this diversity into a single number weakens our grasp of the true picture. It gets further diluted when we take these single numbers from across thousands of employees and condense further into an average to determine our response.  

Yes, averages are easy but also highly misleading. Whenever an average is used to represent an uncertain quantity, it ends up distorting the results because it ignores the impact of the inevitable variations. If there was one movement I’d like to start in the worlds where HR and data intersect, it would be the movement to rethink how we look at averages. 

If not averages, then what?

Think of averages as taking the easy way out. When asked questions and pushed for a quick single number for reference, we usually rely on averages. However, the trend is changing. 

Over the past few years, given the obsession with data, we drifted away from direct employee engagement such as round table discussions, adhoc one to ones and the like. As I find a reverse trend of increased connection circles and leaders leading ‘coffee chats’ and listening circles, the act of supplementing averages with anecdotes is heartening. One wise leader once said, ‘on discovering a conflict between data and anecdotes, always trust the anecdotes.’ Now while the original quote was intended for product feedback, the same can be embraced for most instances. 

But I’d still like some quantitative data

I acknowledge that turning up without any numbers, armed only with anecdotes, can prove detrimental to a conversation especially given the amount of data we are surrounded with. And not once have I said that data isn’t good. It is averages by themselves that I have a beef with. Hence here are my top three alternatives to naked averages. 

Never state the what without the why:

We have a phrase in my team that we use a little too frequently. Used to describe numbers that say very little on their own, ‘naked numbers’ are often a dangerous trap. My feelings for averages (if you didn’t already grasp) are identical to that for naked numbers. Hence my rule of never the what without the why. Even though it annoys some people, when asked for data, I first explain the story on what the number reflects before sharing the number. Doing that forces me to compare against other relevant metrics and supplement with anecdotes. Nowadays when I share numbers without a story, my listeners begin to worry.  

Focus on the distribution, hide the averages:

If you don’t have the luxury of storytelling prior to presenting numbers, here’s an alternate approach. Delete averages altogether. We already use distributions while collecting data and love graphs too. One simple move such as expanding the single average to reflect the response distribution allows the reader to identify the fence sitters and the naysayers. Knowing that a training received an average score of 4+ out of 5 does not tell us much but knowing just how many people responded with what score tells us a wee bit more about how the score ended up being 4+. A distribution over time (e.g. training demand) also allows us to plan resources better. If during the year, for e.g., the average demand for a training was four instances a month but zero in the first and last quarter, distributing resources to facilitate four instances per month does us little good. So, knock the averages off your first page and replace it instead with graphs and distributions. 

I am sure there exists a uniformly divided world where averages make perfect sense. Unfortunately, that world isn’t this one and most likely never will be. If you are crafting strategies for the world we live in, something as easy and opaque as averages is a very inaccurate metric to live by. Just ask those who’ve been tasked to report the average gender pay gap and nothing else. We live in a world riddled with data and it’s time to get smarter on what and how to use what we have. 

So, the next time you hear someone mention averages, share with them the story of the statistician and that you’d like to save the next one from drowning. Let’s rethink that average.