Forum Discussion
Mixpanel API call returns different row counts on desktop and service
I'm calling the Mixpanel endpoint https://data-eu.mixpanel.com/api/2.0/export with the required parameters project id and start/end date.
This works fine on desktop, the data matches what I see when I look at the data in mixpanels own data views.
After uploading the report and configuring the credentials, when I do a refresh on the semantic model it says that it is successful and it does get some data, just not all of it. The data that it does manage to get is spread out over the whole requested date range.
I've been in contact with our IT department responsible for the workspace but they tell me that they can't find any reason for this. The Mixpanel support is not able to help either.
Has anyone here seen something similar to this or know what causes it?
Found the issue myself, it was a combination of a hidden filter and one of the many gotchas with PBI.
TL;DR: There was never an issue with fetching the data.
If you have a text column containing the values "TRUE", "FALSE" and blank(), once you do a data refresh in the service PBI is so kind that it automatically transforms "TRUE" and "FALSE" into -1 and 0. Combine this with a hidden filter where "TRUE" and blank() is selected and voilá, it looks like you are getting fewer rows since you now only see the rows that matches blank().
This behaviour is documented here: https://learn.microsoft.com/en-us/power-bi/connect-data/desktop-data-types#truefalse-type
Oh, and if you want to understand why logical true is represented as -1, start here:
https://stackoverflow.com/questions/23832703/why-does-forth-return-1-as-a-flag-for-true
1 Reply
Found the issue myself, it was a combination of a hidden filter and one of the many gotchas with PBI.
TL;DR: There was never an issue with fetching the data.
If you have a text column containing the values "TRUE", "FALSE" and blank(), once you do a data refresh in the service PBI is so kind that it automatically transforms "TRUE" and "FALSE" into -1 and 0. Combine this with a hidden filter where "TRUE" and blank() is selected and voilá, it looks like you are getting fewer rows since you now only see the rows that matches blank().
This behaviour is documented here: https://learn.microsoft.com/en-us/power-bi/connect-data/desktop-data-types#truefalse-type
Oh, and if you want to understand why logical true is represented as -1, start here:
https://stackoverflow.com/questions/23832703/why-does-forth-return-1-as-a-flag-for-true