more events in customer journey

dbt_project.yml

@@ -4,6 +4,7 @@
 # name or the intended use of these models
 name: 'rudder_sessionization'
 version: '1.0.0'
+config-version: 2 
 
 # This setting configures which "profile" dbt uses for this project.
 profile: 'default'

models/rudder/dbt_aliases_mapping.sql

@@ -13,14 +13,14 @@ Typically a user would login to an application at least once while using the app
         select anonymous_id
           , user_id
           , timestamp as timestamp 
-        from {{ source("<schema>","tracks") }}
+        from {{ source("rudder_1","tracks") }}
 
         union distinct
 
         select user_id
           , null
           , timestamp 
-        from {{ source("<schema>","tracks") }}
+        from {{ source("rudder_1","tracks") }}
 
       )
 

models/rudder/dbt_mapped_tracks.sql

@@ -14,7 +14,7 @@ select *
           ,a2v.dbt_visitor_id
           ,t.timestamp
           ,t.event as event
-        from {{ source("<schema>","TRACKS") }} as t
+        from {{ source("rudder_1","tracks") }} as t
         inner join {{ ref('dbt_aliases_mapping') }} as a2v
         on a2v.alias = coalesce(t.user_id, t.anonymous_id)
         )

models/rudder/dbt_tracks_flow.sql

@@ -18,6 +18,16 @@ with derived_table as (
             nth_value(event,3 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as third_event,
             nth_value(event,4 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as fourth_event,
             nth_value(event,5 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as fifth_event,
+            nth_value(event,6 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as six_event,
+            nth_value(event,7 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as seven_event,
+            nth_value(event,8 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as eight_event,
+            nth_value(event,9 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as ninth_event,
+            nth_value(event,10 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as ten_event,
+            nth_value(event,11 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as eleven_event,
+            nth_value(event,12 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as twelve_event,
+            nth_value(event,13 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as thirteen_event,
+            nth_value(event,14 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as fourteen_event,
+
             from {{ ref('dbt_track_facts') }}
         )
 
@@ -31,4 +41,15 @@ with derived_table as (
             , third_event as event_3
             , fourth_event as event_4
             , fifth_event as event_5
+            , six_event as event_6
+            , seven_event as event_7
+            , eight_event as event_8
+            , ninth_event as event_9
+            , ten_event as event_10
+            , eleven_event as event_11
+            , twelve_event as event_12
+            , thirteen_event as event_13
+            , fourteen_event as event_14
+
+
       from derived_table a

models/rudder/tracks.yml

@@ -1,8 +1,8 @@
 version: 2
 sources:
-    - name: <schema>
+    - name: "rudder_1"
       tables:
-        - name: TRACKS
+        - name: tracks
           columns:
             - name: anonymous_id
 

target/compiled/rudder_sessionization/models/rudder/dbt_aliases_mapping.sql

@@ -0,0 +1,34 @@
+/*
+
+The first intent is to be able to link all events from the same user via some common identifier. Rudder data payload contains two fields anonymous_id and user_id. The first is device specific and generated by the Rudder SDK, the second is the user identifier that can be assigned by application code depending on a user’s registration or login. 
+
+Typically a user would login to an application at least once while using the application on a particular device. The user_id should then become available and be included in event messages. Below code creates a unique identifier dbt_visitor_id that links the user_id to the anonymous_id. The analytic function first_value has been used here
+
+*/
+
+
+
+  with
+      dbt_all_mappings as (
+        select anonymous_id
+          , user_id
+          , timestamp as timestamp 
+        from `rudderstack-367610`.`rudder_1`.`tracks`
+
+        union distinct
+
+        select user_id
+          , null
+          , timestamp 
+        from `rudderstack-367610`.`rudder_1`.`tracks`
+
+      )
+
+      select
+          distinct anonymous_id as alias
+          ,coalesce(first_value(user_id IGNORE NULLS)
+              over(
+                partition by anonymous_id
+                order by timestamp desc
+                rows between unbounded preceding and unbounded following), user_id, anonymous_id) as dbt_visitor_id
+        from dbt_all_mappings

target/compiled/rudder_sessionization/models/rudder/dbt_mapped_tracks.sql

@@ -0,0 +1,20 @@
+/*
+
+ Use the ID generated while creating dbt_aliases_mapping to link all events for the same user on that device. Also note the idle time between events
+
+*/
+
+
+
+select *
+        ,timestamp_diff(cast(timestamp as timestamp), cast(lag(timestamp) over(partition by dbt_visitor_id order by timestamp) as timestamp), minute) as idle_time_minutes
+      from (
+        select t.id as event_id
+          ,t.anonymous_id
+          ,a2v.dbt_visitor_id
+          ,t.timestamp
+          ,t.event as event
+        from `rudderstack-367610`.`rudder_1`.`tracks` as t
+        inner join `rudderstack-367610`.`dbt_jluiscases`.`dbt_aliases_mapping` as a2v
+        on a2v.alias = coalesce(t.user_id, t.anonymous_id)
+        )

target/compiled/rudder_sessionization/models/rudder/dbt_session_duration.sql

@@ -0,0 +1,16 @@
+/*
+
+Table containing a useful session metric - session duration
+
+*/
+
+
+
+select 
+    s1.dbt_visitor_id
+    , s1.session_id
+    , timestamp_diff(cast(s2.ended_at as timestamp), cast(s1.session_start_at as timestamp), minute) as session_duration
+from
+    `rudderstack-367610`.`dbt_jluiscases`.`dbt_session_tracks` as s1
+    LEFT JOIN `rudderstack-367610`.`dbt_jluiscases`.`dbt_session_track_facts` as s2
+      ON s1.session_id = s2.session_id

target/compiled/rudder_sessionization/models/rudder/dbt_session_track_facts.sql

@@ -0,0 +1,17 @@
+/*
+
+With the events mapped to sessions (dbt_session_tracks), one can now stipulate the session end time as the timestamp of the last event (in other words, the highest timestamp) in that session. The next table encapsulates this data. It also contains few columns that can be of aid specifically for analytics involving e-commerce sites/applications.
+
+*/
+
+
+
+
+SELECT s.session_id
+        , cast(MAX(map.timestamp) as timestamp) AS ended_at
+        , count(distinct map.event_id) AS num_pvs
+        , count(case when map.event = 'viewed_product' then event_id else null end) as cnt_viewed_product
+        , count(case when map.event = 'signup' then event_id else null end) as cnt_signup
+      FROM `rudderstack-367610`.`dbt_jluiscases`.`dbt_session_tracks` AS s
+      LEFT JOIN `rudderstack-367610`.`dbt_jluiscases`.`dbt_track_facts` as map on map.session_id = s.session_id
+      GROUP BY 1

target/compiled/rudder_sessionization/models/rudder/dbt_session_tracks.sql

@@ -0,0 +1,15 @@
+/*
+
+A decision is made to treat two events, for the same user, that are separated by 30 minutes or more - as belonging to two different user sessions. The choice of 30 minutes is arbitrary and can be modified as per requirements. Sequence number is assigned to each event within a particular session. Also, the timestamp for the first event in the session is considered as session start time. Start time of the next session is also calculated.
+
+*/
+
+
+
+ select concat(cast(row_number() over(partition by dbt_visitor_id order by timestamp) AS string), ' - ', dbt_visitor_id) as session_id
+      , dbt_visitor_id
+      , timestamp as session_start_at
+      , row_number() over(partition by dbt_visitor_id order by timestamp) as session_sequence_number
+      , lead(timestamp) over(partition by dbt_visitor_id order by timestamp) as next_session_start_at
+from `rudderstack-367610`.`dbt_jluiscases`.`dbt_mapped_tracks`
+where (idle_time_minutes > 30 or idle_time_minutes is null)

target/compiled/rudder_sessionization/models/rudder/dbt_track_facts.sql

@@ -0,0 +1,23 @@
+/*
+
+Below code creates a table to link the track events to the session they belong to. The session association is established via the user identifier linkage and the user session start timestamp. 
+
+So if a user U1 has session S1 with start time as T1 and session S2 with start time as T2 - then event E for user U1 would belong to session S1 if its timestamp falls between T1 and T2 or if T2 is null. The second case occurs for the last recorded session for that user.
+
+*/
+
+
+
+
+select t.anonymous_id
+          , t.timestamp
+          , t.event_id
+          , t.event AS event
+          , s.session_id
+          , t.dbt_visitor_id
+          , row_number() over(partition by s.session_id order by t.timestamp) as track_sequence_number
+        from `rudderstack-367610`.`dbt_jluiscases`.`dbt_mapped_tracks` as t
+        inner join `rudderstack-367610`.`dbt_jluiscases`.`dbt_session_tracks` as s
+        on t.dbt_visitor_id = s.dbt_visitor_id
+          and t.timestamp >= s.session_start_at
+          and (t.timestamp < s.next_session_start_at or s.next_session_start_at is null)

target/compiled/rudder_sessionization/models/rudder/dbt_tracks_flow.sql

@@ -0,0 +1,55 @@
+/*
+
+We leverage analytic functions like first_value and nth_value to create 5-event sequences that capture the flow of events during a session. 5 can be increased or decreased as per requirements.
+
+*/
+
+
+
+with derived_table as (
+          select
+            event_id,
+            session_id,
+            track_sequence_number,
+            first_value(event IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as event,
+            dbt_visitor_id,
+            timestamp,
+            nth_value(event,2 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as second_event,
+            nth_value(event,3 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as third_event,
+            nth_value(event,4 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as fourth_event,
+            nth_value(event,5 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as fifth_event,
+            nth_value(event,6 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as six_event,
+            nth_value(event,7 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as seven_event,
+            nth_value(event,8 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as eight_event,
+            nth_value(event,9 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as ninth_event,
+            nth_value(event,10 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as ten_event,
+            nth_value(event,11 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as eleven_event,
+            nth_value(event,12 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as twelve_event,
+            nth_value(event,13 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as thirteen_event,
+            nth_value(event,14 IGNORE NULLS) over(partition by session_id order by track_sequence_number asc) as fourteen_event,
+
+            from `rudderstack-367610`.`dbt_jluiscases`.`dbt_track_facts`
+        )
+
+          select event_id
+            , session_id
+            , track_sequence_number
+            , event
+            , dbt_visitor_id
+            , cast(timestamp as timestamp) as timestamp
+            , second_event as event_2
+            , third_event as event_3
+            , fourth_event as event_4
+            , fifth_event as event_5
+            , six_event as event_6
+            , seven_event as event_7
+            , eight_event as event_8
+            , ninth_event as event_9
+            , ten_event as event_10
+            , eleven_event as event_11
+            , twelve_event as event_12
+            , thirteen_event as event_13
+            , fourteen_event as event_14
+
+
+      from derived_table a

target/compiled/rudder_sessionization/models/rudder/dbt_user_session_facts.sql

@@ -0,0 +1,17 @@
+/*
+
+Table containing a useful metric about user sessions - no. of sessions for each user
+
+*/
+
+
+
+SELECT
+        dbt_visitor_id
+        , cast(MIN(s.session_start_at) as timestamp) as first_date
+        , cast(MAX(s.session_start_at) as timestamp) as last_date
+        , COUNT(*) as number_of_sessions
+      FROM `rudderstack-367610`.`dbt_jluiscases`.`dbt_session_tracks` as s
+      LEFT JOIN `rudderstack-367610`.`dbt_jluiscases`.`dbt_session_track_facts` as sf
+      ON s.session_id = sf.session_id
+      GROUP BY 1

target/compiled/rudder_sessionization/models/rudder/schema.yml/not_null_dbt_session_duration_session_duration.sql

@@ -0,0 +1,11 @@
+
+
+
+
+
+
+select session_duration
+from `rudderstack-367610`.`dbt_jluiscases`.`dbt_session_duration`
+where session_duration is null
+
+

target/compiled/rudder_sessionization/models/rudder/schema.yml/not_null_dbt_session_tracks_dbt_visitor_id.sql

@@ -0,0 +1,11 @@
+
+
+
+
+
+
+select dbt_visitor_id
+from `rudderstack-367610`.`dbt_jluiscases`.`dbt_session_tracks`
+where dbt_visitor_id is null
+
+

target/compiled/rudder_sessionization/models/rudder/schema.yml/not_null_dbt_session_tracks_session_id.sql

@@ -0,0 +1,11 @@
+
+
+
+
+
+
+select session_id
+from `rudderstack-367610`.`dbt_jluiscases`.`dbt_session_tracks`
+where session_id is null
+
+

target/run/rudder_sessionization/models/rudder/dbt_aliases_mapping.sql

@@ -0,0 +1,45 @@
+
+
+
+
+    create or replace table `rudderstack-367610`.`dbt_jluiscases`.`dbt_aliases_mapping`
+
+
+    OPTIONS()
+    as (
+      /*
+
+The first intent is to be able to link all events from the same user via some common identifier. Rudder data payload contains two fields anonymous_id and user_id. The first is device specific and generated by the Rudder SDK, the second is the user identifier that can be assigned by application code depending on a user’s registration or login. 
+
+Typically a user would login to an application at least once while using the application on a particular device. The user_id should then become available and be included in event messages. Below code creates a unique identifier dbt_visitor_id that links the user_id to the anonymous_id. The analytic function first_value has been used here
+
+*/
+
+
+
+  with
+      dbt_all_mappings as (
+        select anonymous_id
+          , user_id
+          , timestamp as timestamp 
+        from `rudderstack-367610`.`rudder_1`.`tracks`
+
+        union distinct
+
+        select user_id
+          , null
+          , timestamp 
+        from `rudderstack-367610`.`rudder_1`.`tracks`
+
+      )
+
+      select
+          distinct anonymous_id as alias
+          ,coalesce(first_value(user_id IGNORE NULLS)
+              over(
+                partition by anonymous_id
+                order by timestamp desc
+                rows between unbounded preceding and unbounded following), user_id, anonymous_id) as dbt_visitor_id
+        from dbt_all_mappings
+    );
+