Seeking Methods for More Informative Median in Frequency Analysis

·

does anyone know if there's a way to obtain a more informative median (or something similar) when calculating frequency? as it stands, frequency is always an integer, so the median will always be an integer, often resulting in graphs like this one where small shifts in the distribution don't show up; I can use the average, of course, but that tends to get skewed by outliers...

🎫1

5 comments

Thena
APP
·
·
Hey Vlad! Julianne from Mixpanel Support here. Would the Frequency per User Behavioral property work here? While not the median, it would allow you to see more of those small shifts in distribution. Another option would be to graph the P25, P75, and P99 percentiles on the same chart. This would also let you see more of those shifts that may be missed when looking only at the median.
🤔1
Vlad S.
·
·
somewhat related question: in a funnel report, how can I calculate the median/average frequency per user? I can break it down like this and you can sort of see that the distributions are different, but it's very hard to read or do much with it
Thena
APP
·
·
Hey Vlad! I hate to be the bearer of bad news, but the median/average frequency per user for a Funnel report is a product gap on our side. That said, I think this would be a great functionality to add to the product. I definitely recommend sending your use case through the 'Submit Feedback' form (under the ? menu), which will flag it to the Product team.
👍1
Vlad S.
·
·
yeah, I think in general a "histogram plot" would be a good addition alongside line and bar plots
Vlad S.
·
·
it's almost a third of what I end up building in Mixpanel, and it would be super helpful to have a visualization type specifically designed for it rather than have to hack around it using bar plots