Expand my Community achievements bar.

Join us at Adobe Summit 2024 for the Coffee Break Q&A Live series, a unique opportunity to network with and learn from expert users, the Adobe product team, and Adobe partners in a small group, 30 minute AMA conversations.

Medians/percentile in calculated metrics for Rows

Avatar

Level 4

3/1/19

Median and percentile functions are available as built-in functions for calculated metrics within Adobe Analytcs, but they are Table based,

From Adobe Documentation:

Table Functions versus Row Functions

A table function is one where the output is the same for every row of the table. A row function is one where the output is different for every row of the table.

See below example of table using the this functionality with one Median and one 90th Percentile metric.

percentile.png

The base metric that is calculated as median and 90th percentile is a Numeric metric. With Row capability for median and percentile, instead of having the median/percentile for  all line items in total it would be possible to see the median for each individual row. Ie, what was the median/percentile for Feb 4, Feb 11, Feb18.

This functionality would be fantastic and help to calculate medians/percentiles directly in the Adobe interface without having to resort to large data exports and manual calculation. Useful for any Numeric metrics where you need to understand the distribution of the individual metric values (eg technical data such as load times, waiting times, or other numeric metrics such as sales numbers) across different dimension values.

Below question experiences the same issue as me.

Median function in calculated metrics

4 Comments

Avatar

Level 1

7/25/19

This is a really cool idea. The same goes for standard deviation or t-test. As they are table based at the moment, their use is very limited. If they were row based, Analytics would have much more value for us.

Avatar

Employee

11/8/19

I understand how row-based percentiles would be useful.  And being able to see the standard deviation could also be useful, but how would you use t-test?  What would you be testing?