# Designing eLearning for complex topics: averages vs percentiles.

Hi everyone. I've been tasked with coming up with a way of explaining revised metrics for performance within our large call centre. One  of the major changes is a shift away from averages, and toward percentiles - for instance, when measuring call times.  While I'm sure it's a very basic topic for statisticians, it's surprisingly difficult to explain in a clear and concise way.

Has anyone got any creative and effective ideas for how I can explain the difference between averages and percentiles?

###### 7 Replies

Hi Erica,

Could you start by explaining to me what the difference is? Then I can help you pare it down to something clear and concise.

It may also be helpful to explain why this change is being implemented.

I'd use a concrete example to explain the concept. For instance, you could show 10 different people of varying heights. I'd add an animation draw a line across their heads showing where the average would be (let's say 5'8"). Then, you could focus in on the 2nd tallest person and say "Dave is 6'1", but Seth is taller. In this group of people, Dave's height is in the 90th percentile. He's taller than 90% of the group."

You could then take Dave and put him in with a taller or shorter group of people to show how his percentile changes when the overall group changes. You could also compare the 50th percentile to the average to show they're not the same. When you've finished the concrete example you can translate it back to call metrics. Does that make sense?

Thanks Allison.  So previously our benchmarks have been calculated as averages - so adding together the call times for all the calls taken in a month (for example) and dividing that number by the total number of calls to come up with an average call time.

The main issue with that way of determining benchmarks is that this number can be skewed by outliers (extremely long or short calls) so may not be a really accurate representation of someone's performance.

To calculate a percentile, in this case, the calls for the month would be ordered from shortest to longest, and then a call at the 80th percentile (the point at which 80% of the call times are below, and 20% above) would become the measurement for the call-taker.  This measurement, potentially could be a more accurate metric for a call-taker's performance.

That explanation is very clear. Could you simply add some graphics to show examples say 20 calls and then the different outcomes so the learner can visually see the difference?

Thanks very much Tristan. That's what I've got with in the end - an animated representation of how an average is calculated from 10 calls, and then how a percentile would be calculated from the same 10 calls.

Hi Erica,

So I totally agree with my fellow IDs that visual is the way to go here. Use bar graphs, not numbers for example.

One other thought... Step back from the math a bit and think about what you are trying to accomplish. If before you were measuring everyone against the mid point (ie average) and now you are setting the bar at the 80th percentile, show just that... just that. Forget the words "average" and "percentile" altogether as they have no bearing for the learner.   Instead just visually show the difference between measuring one way and then the other - one set of bar graphs with the two different metrics called out as vertical arrows for example..

Then have them apply the new way of measuring success.... "Order the following call times from low to high and indicate what the new performance standard would be for this set of calls" (Note: Have 10 bars and place a colored indicator underneath the space where the 8th bar would go)

The bottom line is that it sounds like you don't need everyone to be able to do the math, but to understand how (and why!) the benchmark has changed.  Just teach the math piece to the folks that really need it.

Hope this helps!