How to build a closed funnel of user steps in Big Query? - google-bigquery

Question

Please help me with the BigQuery query. I need to build a closed funnel of user steps events in a mobile app for a week.
The table looks like this:
It is necessary to collect all unique users who have passed from step 1 to step 2 and so on to step 6 during this period. Between these steps, they could do something else, be distracted by other events. But what is important is the passage of each unique user through these steps in a given period of time.
Please tell me how to create such a funnel?

Answer 1

There can be multiple ways of achieving this. Here is an approach using identical sample data, which is not the most optimal but is very self-explanatory and definite:
with data as (
select 'a' as user_id, cast('2020-01-01 04:45:00' as timestamp) as event_timestamp, '1' as step_name
union all
select 'b' as user_id, cast('2020-01-01 04:50:00' as timestamp) as event_timestamp, '1' as step_name
union all
select 'a' as user_id, cast('2020-01-01 05:00:00' as timestamp) as event_timestamp, '2' as step_name
union all
select 'a' as user_id, cast('2020-01-01 05:15:00' as timestamp) as event_timestamp, '3' as step_name
union all
select 'b' as user_id, cast('2020-01-01 04:55:00' as timestamp) as event_timestamp, '2' as step_name
union all
select 'c' as user_id, cast('2020-01-01 04:58:00' as timestamp) as event_timestamp, '1' as step_name
union all
select 'a' as user_id, cast('2020-01-01 05:16:00' as timestamp) as event_timestamp, '4' as step_name
union all
select 'b' as user_id, cast('2020-01-01 05:16:00' as timestamp) as event_timestamp, '3' as step_name
),
data2 as (
select a.user_id, a.step_name step_1, b.step_name step_2, c.step_name step_3, d.step_name step_4 from ( select user_id, event_timestamp, step_name from data where step_name = '1') a
left join data b on (a.user_id = b.user_id and a.event_timestamp < b.event_timestamp and b.step_name = '2')
left join data c on (b.user_id = c.user_id and b.event_timestamp < c.event_timestamp and c.step_name = '3')
left join data d on (c.user_id = d.user_id and c.event_timestamp < d.event_timestamp and d.step_name = '4')
)
select * from (
select 'step_1' as event_name, count(distinct user_id) as n_users from data2 where step_1 is not null
group by 1
union all
select 'step_2' as event_name, count(distinct user_id) as n_users from data2 where (step_1 is not null and step_2 is not null)
group by 1
union all
select 'step_3' as event_name, count(distinct user_id) as n_users from data2 where (step_1 is not null and step_2 is not null and step_3 is not null)
group by 1
union all
select 'step_4' as event_name, count(distinct user_id) as n_users from data2 where (step_1 is not null and step_2 is not null and step_3 is not null and step_4 is not null)
group by 1
)
order by 1
You can further optimize this based on your specific filters, conditions, etc.