Hi @mrv
Great question, it’s definitely possible to write this query without CTEs. If you run your query you’ll see an error:
ERROR: column "ab_test_name" does not exist LINE 10: AND ab_test_name = 'longer_onboarding_201803' ^
This is because you can’t reference a column (column "ab_test_name" does not exist
) you just defined inside the WHERE clause. Let’s fix it:
SELECT
custom_parameters ->> 'ab_test_name' AS ab_test_name,
custom_parameters ->> 'ab_test_variation' AS ab_test_variation,
COUNT(DISTINCT(user_id))
FROM mobile_analytics.events
WHERE
custom_parameters ->> 'ab_test_name' IS NOT NULL
AND custom_parameters ->> 'ab_test_name' = 'longer_onboarding_201803'
AND action = 'signup'
GROUP BY 1, 2
I’ve removed categorized_at
column, I guess it made there by accident. We want to know how many users we had per AB-test variation, no need to pull in this datetime column.
Note how I referenced columns inside GROUP BY
by their numbers – 1, 2
. It’s simply to avoid duplication and if you want to group by another column (say, country
) you can just change it in the SELECT
.
As you can see, this query seems very bulky, CTEs allow you to save some duplication and make it more readable.
Ping me if you have more questions