Schematized Snowflake Pipeline
This guide describes how Mixpanel data is exported into a Snowflake (opens in a new tab) dataset. Create a pipeline (opens in a new tab) to export your Mixpanel data into Snowflake. Once an export job is scheduled, Mixpanel exports data to Snowflake on a recurring basis.
Design
Mixpanel exports data to its own Snowflake account and gives your Snowflake account access to read the data. As a result, you will need a Snowflake account to be able to use the exported data. For more information on how to use your shared data, see Sharing Data in Snowflake (opens in a new tab).
For Snowflake export, we first load the data into a single-column raw (VARIANT type) data table. This data will be a transformed version of the raw data stored in Mixpanel. Then, we create a view to expose all properties as columns. The view name is the cleaned version of the event name and the raw table name is always the view name suffixed with _raw
. For example, for signup
event, you will have a signup_raw
table and a signup
view. Please see Schema for general information about the schemas in Schematized Export Pipelines.
Partitioning
The data in the raw tables is clustered based on time
column but in project's timezone. To be exact, we use CLUSTER BY (TO_DATE(CONVERT_TIMEZONE('UTC','<TIMEZONE>', TO_TIMESTAMP(DATA:time::NUMBER)))
where TIMEZONE
is the Mixpanel project's timezone.
Queries
Mixpanel recommends you place all events into a single table to make querying easier. To get more information about the table schemas, please see Schema.
A query is a request for data results. You can perform actions on the data, such as combine data from different tables; add, change, or delete table data; and perform calculations.
Snowflake supports a VARIANT type that can store JSON objects and arrays. Mixpanel exposes array and object top-level properties as VARIANT columns in the view.
Here is an example of how you can query the raw table when using one table for all the events:
SELECT count(*)
FROM mixpanel.mp_master_event_raw
WHERE data:mp_event_name::string = “Signup”;
Here is an example of how you can query the view when using one table for all the events:
SELECT count(*)
FROM mixpanel.mp_master_event
WHERE mp_event_name = “Signup”;
Here is an example of how you can query the raw table when using multiple tables for the events:
SELECT count(*)
FROM mixpanel.signup_raw
WHERE data:distinct_id::string = “1”;
Here is an example of how you can query the view when using multiple tables for the events:
SELECT count(*)
FROM mixpanel.signup
WHERE distinct_id = “1”;
Getting the number of events in each day
You will need this if you suspect the export process is not exporting all the events you want. As time column in the tables is in UTC timezone, you first need to convert that to your Mixpanel project timezone, and then, get the number of events for each day. The following query will do that for you.
SELECT TO_DATE(CONVERT_TIMEZONE('UTC','<PROJECT_TIMEZONE>', time)) as ttime, count(*)
FROM "<DB_NAME>"."PUBLIC"."MP_MASTER_EVENT"
WHERE ttime>=TO_DATE('2021-12-03') AND ttime<=TO_DATE('2021-12-07')
GROUP BY ttime
This example returns the number of events in each day in project timezone for a monoschema export pipeline and an example daterange. PROJECT_TIMEZONE
and DB_NAME
should be replaced by your Mixpanel project timezone and your snowflake database name. You can adjust the query for multischema by putting the right table name in the query.
Querying the identity mapping table
When using the ID mappings table, you should use the resolved distinct_id
in place of the non-resolved distinct_id
whenever present. If there is no resolved distinct_id
, you can then use the distinct_id
from the existing people or events table.
Below is an example SQL query that references the ID mapping table to count number of events in a specific date range for each unique user in San Francisco
SELECT
CASE
WHEN m.resolved_distinct_id IS NOT NULL THEN m.resolved_distinct_id
WHEN m.resolved_distinct_id IS NULL THEN e.distinct_id
END as resolved_distinct_id,
COUNT(*) AS count
FROM "<DB_NAME>"."PUBLIC"."MP_MASTER_EVENT" e FULL OUTER JOIN "<DB_NAME>"."PUBLIC"."MP_IDENTITY_MAPPINGS_DATA" m
ON e.distinct_id = m.distinct_id
AND mp_city="San Francisco"
AND TO_DATE(CONVERT_TIMEZONE('UTC','<PROJECT_TIMEZONE>', e.time)) >= TO_DATE("2020-04-01")
AND TO_DATE(CONVERT_TIMEZONE('UTC','<PROJECT_TIMEZONE>', e.time)) <= TO_DATE("2020-04-03")
GROUP BY resolved_distinct_id
LIMIT 100
Counting number of times a user has done a specific behavior is also possible by adding more filters on event properties. You can adjust the query for multischema by putting the right table name in the query.
Was this page useful?