When counting total users in Firebase Console and Bigquery export for project, we have a massive difference in users count:
OS | Firebase (Total Users) | BigQuery (Users) | Difference |
---|---|---|---|
Windows (Web) | 38k | 23k | 39% |
Android | 21k | 24k | 15% |
Ios | 10k | 12k | 20% |
other platforms (Web) | 1,5k | 1k | ~50% |
Total | 70k | 60k | 15% |
BigQuery query:
SELECT
count(distinct user_pseudo_id),
device.operating_system
FROM
`{project}.events_*`
where _TABLE_SUFFIX between "20210614" and "20210620"
GROUP BY
device.operating_system;
I've read about Firebase approximation and HLL, but it shouldn't give that much difference?
Both platorms have filter on date, so it queries 1 week old data (14 june - 20 june).
Any suggestions, what can cause this difference and how to fix it?