Get cohort analysis.
If you specify neither an interval nor a unit, the interval is 1 day. This means that each user gets 24 hours in each interval to do the specified event.
An example response with a born_event of 'event integration' and event of 'viewed report' might look like this:
{
"2012-01-01": {
"counts": [2, 1, 2],
"first": 2
},
"2012-01-02": {
"counts": [9, 7, 6],
"first": 10
},
"2012-01-03": {
"counts": [9, 6, 4],
"first": 10
}
}
These results indicate that on 2012-01-02, 10 users did the born_event ("event integration"), as indicated by the first field. If the retention_type=compounded, then first will instead indicate the number of users who did event ("viewed report") on the specified date. 9 of those users did event ("viewed report") within 24 hours (the "0th" interval) of doing the born_event. Seven of those did event between 24 and 48 hours (interval 1) of the born_event. These 7 are a subset of the initial 10, but not necessarily a subset of the 9 (retention is not a funnel; see the number increase between 72 and 96 hours). And finally, 6 users did event between 48 and 72 hours (interval 2) after the born_event.
In the Mixpanel retention UI, "First time" corresponds to retention_type=birth, and "Recurring" corresponds to retention_type=compounded.
The Query API has a rate limit of 60 queries per hour and a maximum of 5 concurrent queries.