Understanding Metric Report Formula Differences from Line Report

Ina G. · 2024-11-07 09:06:59.036212+00

Hi everyone again :slightly_smiling_face: can anyone please explain what is the exact formula used to calculate the overall number in a *Metric report* , please? the reason I am asking is that using the same information, but in the *Line report*, it shows really different numbers..

Understanding Metric Report Formula Differences from Line Report | Mixpanel Community

14 comments

· Sorted by Oldest

Andrew S.
·
·
Hard to say without knowing more about the insight you are running. Often the difference is between “over time” which may have averages at different points and then an overall average where the metric is then just the average now or at a specific time based on the measured time period. Like I said, hard to be more helpful without knowing specifically is being measured, time periods, etc.
Ina G.
·
·
sure I am trying to see how many suggestions are added on average to posts (the event is "Add a suggestion") by power, core and casual users
Andrew S.
·
·
So are you doing a measure of total on three metrics each filtered by a different cohort? Or are you measuring an aggregate property per user? Etc.
Andrew S.
·
·
E.g
Andrew S.
·
·
This metric is looking at total events of four time periods (so each metric has a filter applied based on time)
Loading attachments
Andrew S.
·
·
But looking over time shows very different numbers
Loading attachments
Ina G.
·
·
Loading attachments
Andrew S.
·
·
That is because the line over time is looking at how many total events on each of the days meet the criteria for the specific metric. So the 7 day will have the same numbers as the other three each day. But by day 8 it will show 0 where they will not. The 14 day will show until day 15, etc
Andrew S.
·
·
But a line is showing me daily values. And an average daily value. The metric is showing me the sum of the daily values (based on my measured metric)
Andrew S.
·
·
Loading attachments
Andrew S.
·
·
So I think you are seeing different results bc of the measurement basis. For example, when you do a line chart you are measuring hourly or daily or weekly, etc. and that will produce different results. With a metric, you do not pick a sub period like that, so I believe it is looking at overall the time period.
💯1
Andrew S.
·
·
When I tried to mimic your metrics, I get a higher number for my p90 than if I look at the weekly line bc over 3 months the p90 is X where for a weekly average measurement over a 3 mos period, it is Y. So not really the same thing.
Loading attachments
Ina G.
·
·
I think I understand the logic behind: With a metric, you do not pick a sub period like that, so I believe it is looking at overall the time period but since it looks at the "cumulative total number" over a period of months (like the amount from February + amount from March + etc.) shouldn't the overall number be bigger then?
Loading attachments
Andrew S.
·
·
If it were a sum and not an average maybe. But the average or percentiles can be skewed by particular measurement periods
🙌1

Andrew S.
·
·
Hard to say without knowing more about the insight you are running. Often the difference is between “over time” which may have averages at different points and then an overall average where the metric is then just the average now or at a specific time based on the measured time period. Like I said, hard to be more helpful without knowing specifically is being measured, time periods, etc.
Ina G.
·
·
sure I am trying to see how many suggestions are added on average to posts (the event is "Add a suggestion") by power, core and casual users
Andrew S.
·
·
So are you doing a measure of total on three metrics each filtered by a different cohort? Or are you measuring an aggregate property per user? Etc.
Andrew S.
·
·
E.g
Andrew S.
·
·
This metric is looking at total events of four time periods (so each metric has a filter applied based on time)
Loading attachments
Andrew S.
·
·
But looking over time shows very different numbers
Loading attachments
Ina G.
·
·
Loading attachments
Andrew S.
·
·
That is because the line over time is looking at how many total events on each of the days meet the criteria for the specific metric. So the 7 day will have the same numbers as the other three each day. But by day 8 it will show 0 where they will not. The 14 day will show until day 15, etc
Andrew S.
·
·
But a line is showing me daily values. And an average daily value. The metric is showing me the sum of the daily values (based on my measured metric)
Andrew S.
·
·
Loading attachments
Andrew S.
·
·
So I think you are seeing different results bc of the measurement basis. For example, when you do a line chart you are measuring hourly or daily or weekly, etc. and that will produce different results. With a metric, you do not pick a sub period like that, so I believe it is looking at overall the time period.
💯1
Andrew S.
·
·
When I tried to mimic your metrics, I get a higher number for my p90 than if I look at the weekly line bc over 3 months the p90 is X where for a weekly average measurement over a 3 mos period, it is Y. So not really the same thing.
Loading attachments
Ina G.
·
·
I think I understand the logic behind: With a metric, you do not pick a sub period like that, so I believe it is looking at overall the time period but since it looks at the "cumulative total number" over a period of months (like the amount from February + amount from March + etc.) shouldn't the overall number be bigger then?
Loading attachments
Andrew S.
·
·
If it were a sum and not an average maybe. But the average or percentiles can be skewed by particular measurement periods
🙌1