It can be tricky to decide which aggregator to use, especially when it comes to average and median. You can always use both of them - one metric to calculate the average and another the median. This lets you see which measure will be most useful in a particular case. However, understanding these statistical terms will make it easier to choose the right aggregator.
Average (or mean) and median play the similar role in understanding the central tendency of a set of numbers. Average has traditionally been a popular measure of a middle point in a set, but it has a disadvantage of being influenced by single values which are much higher or lower than the rest of the values. That’s why the median is a better midpoint measure for cases where a small number of outliers could drastically skew the average.
Average Median
Definition |
The average is the arithmetic mean of a set of numbers. |
The median is a numeric value that separates the higher half of a set from the lower half. |
When is it applicable? |
The mean is used for normal number distributions, which have a low amount of outliers. |
The median is generally used to return the central tendency for skewed number distributions. |
How is it calculated? |
The average is calculated by adding up all the values and dividing the sum by the total number of values. |
The median can be calculated by listing all numbers in ascending order and then locating the number in the centre of that distribution. |
Example: Normal distribution |
2, 3, 3, 5, 8, 10, 11 (2+3+3+5+8+10+11)/7= 6 AVG = 6 |
2, 3, 3, 5, 8, 10, 11
MED = 5 |
Example: Skewed distribution |
2, 2, 3, 3, 5, 7, 8, 130 (2+2+3+3+5+7+8+130)/8= 20 AVG = 20 |
2, 2, 3, 3, 5, 7, 8, 130 (3+5)/2=4 MED = 4 |
Conclusion
If the data you are comparing is mostly uniform then you can safely use the average (AVG) aggregator. However, if your number set has some outliers then you need to consider using median (MED) to filter out the values that are skewing the results.
Examples
A few practical examples:
- In order to report on full resolution time, use the Explore Full Resolution Time (hrs) [MED] metric. Choose the median operator because a number of tickets have been under investigation for a while and these tickets might skew your report.
- To check the average amount of replies posted by the agents use the # Replies [MED] metric because the number of replies is more or less constant.
- To figure out how fast the support team replies to new requests, use the Explore First Reply Time (hrs) [MED] metric. Since first reply time is normally constant, create a metric that will count the average first reply time. Additionally, you can filter out proactive tickets, created by agents from the report because they might have a high first reply time.
8 Comments
Hi Eugene,
You mention
"I will filter out proactive tickets, the once created by agents, from the report because most of them have irregularly high First reply time."
How do you filter out pro-active tickets (agent created tickets) from your first reply time avg tickets?
I've been looking for a way to do this for quite some time.
Is it simply filtering out 'ticket created by [agent]' ?
----
[Edit] I just found your great post explaining exactly this question. Sorry about that!
Looking forward to finding out how.
Hi Breydon,
The easiest way to filter out the proactive tickets would be by using the Ticket Submitter Role attribute, which unfortunately is not available in Insights. It is available in our new reporting product - Explore, that is in the beta stage at the moment.
The best workaround in Insights is to use the Ticket Submitter filter. Since you want to exclude all tickets submitted by agents the filter configuration will be: Ticket Submitter isn't (names of your agents).
is there a way to be able to set up quartiles in reports? For example, if the median is a numeric value that separates the higher half of a set from the lower half. What formula do i use in order to figure out the median value for the lower half and hight half of the data set?
Hi Madalina,
I'm not as familiar with the mathematical principles that you are trying to work through, but if you are familiar with how to generate the equations outside of GoodData, you may be able to use GoodData's Help Center Sections on Metrics and MAQL to help build a query out that accomplishes this for you.
I will call out that the help center may not be specific to Zendesk, but for an overview of you you might use GoodData as a platform, I'd recommend taking a look to see if something clicks for you.
hi,
The Selecting metric aggregators. link doesn't work.
Could you please indicate me where I can find a guide to making one queries and reports in Explore.
Hey Caroline,
Thanks for sharing!
I'm going to pass this information to our Documentation team to see if we can get that link updated to the correct URL.
Cheers!
Hi Caroline, I've now fixed the link. Many thanks for pointing this out to us!
On Wednesday, April 8, from 8:00am-10:00pm CDT, we'll have a special guest team of experts on hand to answer your questions about Explore: Reporting & Insights Migration in your Zendesk Explore instance for an AMA-style conversation!
Just click the "new post" button and write up your question. Be sure to be as detailed as possible. The best questions include:
Questions posted to this topic in advance of the start time will be answered first thing on the 8th, so if you've got something you want our experts to look at, post it today!
Please sign in to leave a comment.