Group events by percentile

[OscarMM]

Hi community,

I need to know how many users are repeating an event N times during a range of time by percentile. For example:

• 25% of users triggered event 12 times.
• 50% of users triggered event 8 times.
• 75% of users triggered event 3 times.

I can get percentage using calculated metrics, something like this < Unique User(Ocurrences) / All Users(Ocurrences) >.

But how can I group them by percentile. I tought in sequencial segment, but I think we can't use calculated metrics in segment builder.

Any clue?

Thank you.

[Jennifer_Dungan · Accepted solution]

Well it's not really a percentile, this is more like % of users that perform the action x times, y times, or z times... but you let's say you have an event (let's use event1 for this example), and that event is trigger on your action.

You can create segment buckets for actions performed in a set time.

So if you want to look at actions per Visit, you can create buckets that look like:

[Visit Level Container]

event1 is less than or equal to 5

---

[Visit Level Container]

event1 is greater than or equal to 6

AND

event1 is less than or equal to 10

---

[Visit Level Container]

event1 is greater than or equal to 11

AND

event1 is less than or equal to 20

---

etc....

If you don't want to use "Visit" but say something like the last 30 full days, then you can use:

---

@OscarMM wrote:

[Visitor Level Container]

event1 is less than or equal to 5

AND

Last 30 Full Days

---

[Visitor Level Container]

event1 is greater than or equal to 6

AND

event1 is less than or equal to 10

AND

Last 30 Full Days

---

etc

(don't forget to have a last catch-all of greater than or equal to X to catch the upper scale (just like the lowest bucket is just less than or equal to)

You can change the bucket ranges to suit your needs, but this will create non-overlapping buckets based on the number of times users performed your action. I assume you will actually be using the second option (Visitor level segments, since you want to pair with UVs, and adding the date range to these segments ensure you aren't looking at Visitor data for all time - I highly recommend that you have your panel date range match the date range used in the segment to prevent oddities - particularly since your buckets will be date bound, but your total UVs in the coming steps will not)

Now, you can create a freeform table with a column of ALL UVs, then you can create Columns for each of your buckets using the same metric (UVs):

From here, you can now create adhoc calculated metrics using the main UV column and each individual column to create a percentage of users:

Select the to columns, right click, "Create metric from selection" > "Divide"

You will need to modify the calculated metric to ensure the order of the division is correct, you are displaying it as %, and you are showing the appropriate number of decimals for your desired display (and that you name it appropriately):

Once you have your percentages, you don't actually have to keep the other columns if you don't want.. or you can keep them for reference.