updated bqml to use clientId vs fullvisitorid and removed unused files

llooker · Jul 27, 2020 · 5ea5056 · 5ea5056
1 parent 433f89a
commit 5ea5056
Show file tree

Hide file tree

Showing 6 changed files with 48 additions and 126 deletions.
diff --git a/BQML/bqml.model.lkml b/BQML/bqml.model.lkml
diff --git a/BQML/predictions.view.lkml b/BQML/predictions.view.lkml
@@ -42,7 +42,7 @@ view: future_purchase_model {
     --, CLASS_WEIGHTS=[('1',1), ('0',0.05)] -- Consider adding class weights or downsampling if you have imbalanced classes
     ) AS
     SELECT
-    * EXCEPT(fullVisitorId)
+    * EXCEPT(clientId)
     FROM ${training_input.SQL_TABLE_NAME};;
   }
 }
@@ -82,7 +82,7 @@ view: roc_curve {
     type: number
     link: {
       label: "Likely Customers to Purchase"
-      url: "/explore/bqml_ga_demo/ga_sessions?fields=ga_sessions.fullVisitorId,future_purchase_prediction.max_predicted_score&f[future_purchase_prediction.predicted_will_purchase_in_future]=%3E%3D{{value}}"
+      url: "/explore/bqml_ga_demo/ga_sessions?fields=ga_sessions.clientId,future_purchase_prediction.max_predicted_score&f[future_purchase_prediction.predicted_will_purchase_in_future]=%3E%3D{{value}}"
       icon_url: "http://www.looker.com/favicon.ico"
     }
   }
@@ -153,9 +153,9 @@ view: future_input {
   extends: [user_facts]
   derived_table: {
     sql: {% assign x  = "${EXTENDED}" %}
-    {% assign updated_start_sql = x | replace: 'DAYS_BACK',"6"   %}
+    {% assign updated_start_sql = x | replace: 'DAYS_BACK',"30"   %}
     /*updated_start_date*/
-    {% assign updated_sql = updated_start_sql  | replace: 'DAYS_FROM',"7"  %}
+    {% assign updated_sql = updated_start_sql  | replace: 'DAYS_FROM',"31"  %}
      /*updated_end_date*/
     {{updated_sql}}
     ;;
@@ -165,7 +165,7 @@ view: future_input {
 
 view: future_purchase_prediction {
   derived_table: {
-    sql: SELECT fullVisitorId,
+    sql: SELECT clientId,
           pred.prob as user_propensity_score,
           NTILE(10) OVER (ORDER BY pred.prob DESC) as user_propensity_decile
         FROM ml.PREDICT(
@@ -177,10 +177,10 @@ view: future_purchase_prediction {
   }
   dimension: user_propensity_score {type: number}
   dimension: user_propensity_decile {type: number}
-  dimension: fullVisitorId {
+  dimension: clientId {
        type: string
       hidden: no
-      sql: TRIM(REPLACE(${TABLE}.fullVisitorId,',','')) ;;
+      sql: TRIM(REPLACE(${TABLE}.clientId,',','')) ;;
       }
   measure: average_user_propensity_score {
     type: average
@@ -190,4 +190,4 @@ view: future_purchase_prediction {
     type:  average
     sql:  ${user_propensity_decile} ;;
   }
-}
+}
diff --git a/BQML/user_facts.view.lkml b/BQML/user_facts.view.lkml
@@ -7,31 +7,31 @@ view: user_facts {
         WHERE TIMESTAMP(PARSE_DATE('%Y%m%d', REGEXP_EXTRACT(_TABLE_SUFFIX,r'\d\d\d\d\d\d\d\d')))  BETWEEN ((TIMESTAMP_ADD(TIMESTAMP_TRUNC( CURRENT_TIMESTAMP(), DAY), INTERVAL -DAYS_BACK DAY))) AND ((TIMESTAMP_ADD(TIMESTAMP_ADD(TIMESTAMP_TRUNC(CURRENT_TIMESTAMP(), DAY), INTERVAL -DAYS_BACK DAY), INTERVAL DAYS_FROM DAY)))),
       -- labeling customers who have made a purchase as a 1 and customers who have not made a purchas as a 0
     user_label AS (
-      SELECT fullvisitorId, max(case when totals.transactions >= 1 then 1 else 0 end) as label, max(case when totals.transactions >= 1 then visitStartTime end) as event_session
+      SELECT clientId, max(case when totals.transactions >= 1 then 1 else 0 end) as label, max(case when totals.transactions >= 1 then visitStartTime end) as event_session
         FROM filtered_base
-        GROUP BY fullvisitorId),
+        GROUP BY clientId),
       -- finding the most common hour of day for each user within the time period
     unique_hour_of_day AS(
-      (SELECT   ga_sessions_visit_start_hour_of_day, fullVisitorId FROM (SELECT ROW_NUMBER () OVER(PARTITION BY fullVisitorId ORDER BY   pageviews) as row_number, fullVisitorId, ga_sessions_visit_start_hour_of_day
-      FROM (SELECT ga_sessions.fullVisitorId as fullvisitorid, EXTRACT(HOUR FROM TIMESTAMP_SECONDS(ga_sessions.visitStarttime)) AS ga_sessions_visit_start_hour_of_day, SUM(ga_sessions.totals.pageviews) as pageviews
-      FROM filtered_base  AS ga_sessions  LEFT JOIN user_label ON ga_sessions.fullvisitorid = user_label.fullvisitorid  @{QUERY_FILTER}  GROUP BY 1,2)) WHERE row_number = 1)),
+      (SELECT   ga_sessions_visit_start_hour_of_day, clientId FROM (SELECT ROW_NUMBER () OVER(PARTITION BY clientId ORDER BY   pageviews) as row_number, clientId, ga_sessions_visit_start_hour_of_day
+      FROM (SELECT ga_sessions.clientId as clientId, EXTRACT(HOUR FROM TIMESTAMP_SECONDS(ga_sessions.visitStarttime)) AS ga_sessions_visit_start_hour_of_day, SUM(ga_sessions.totals.pageviews) as pageviews
+      FROM filtered_base  AS ga_sessions  LEFT JOIN user_label ON ga_sessions.clientId = user_label.clientId  @{QUERY_FILTER}  GROUP BY 1,2)) WHERE row_number = 1)),
       -- findiing the most common metro for the user within the time period
       unique_dma AS(
-      (SELECT   metro, fullVisitorId FROM (SELECT ROW_NUMBER () OVER(PARTITION BY fullVisitorId ORDER BY   pageviews) as row_number, fullVisitorId, metro
-      FROM (SELECT ga_sessions.fullVisitorId as fullvisitorid, ga_sessions.geoNetwork.metro as metro , SUM(ga_sessions.totals.pageviews) as pageviews
-      FROM filtered_base  AS ga_sessions LEFT JOIN user_label ON ga_sessions.fullvisitorid = user_label.fullvisitorid @{QUERY_FILTER}  GROUP BY 1,2)) WHERE row_number = 1)),
+      (SELECT   metro, clientId FROM (SELECT ROW_NUMBER () OVER(PARTITION BY clientId ORDER BY   pageviews) as row_number, clientId, metro
+      FROM (SELECT ga_sessions.clientId as clientId, ga_sessions.geoNetwork.metro as metro , SUM(ga_sessions.totals.pageviews) as pageviews
+      FROM filtered_base  AS ga_sessions LEFT JOIN user_label ON ga_sessions.clientId = user_label.clientId @{QUERY_FILTER}  GROUP BY 1,2)) WHERE row_number = 1)),
       -- finding the most common day of week for the user within the time period
       unique_day_of_week AS(
-      (SELECT   ga_sessions_visit_start_day_of_week, fullVisitorId FROM (SELECT ROW_NUMBER () OVER(PARTITION BY fullVisitorId ORDER BY   pageviews) as row_number, fullVisitorId, ga_sessions_visit_start_day_of_week
-      FROM (SELECT ga_sessions.fullVisitorId as fullvisitorid, FORMAT_TIMESTAMP('%A', TIMESTAMP_SECONDS(ga_sessions.visitStarttime)) AS ga_sessions_visit_start_day_of_week  , SUM(ga_sessions.totals.pageviews) as pageviews
-      FROM filtered_base  AS ga_sessions LEFT JOIN user_label ON ga_sessions.fullvisitorid = user_label.fullvisitorid  @{QUERY_FILTER} GROUP BY 1,2)) WHERE row_number = 1)),
+      (SELECT   ga_sessions_visit_start_day_of_week, clientId FROM (SELECT ROW_NUMBER () OVER(PARTITION BY clientId ORDER BY   pageviews) as row_number, clientId, ga_sessions_visit_start_day_of_week
+      FROM (SELECT ga_sessions.clientId as clientId, FORMAT_TIMESTAMP('%A', TIMESTAMP_SECONDS(ga_sessions.visitStarttime)) AS ga_sessions_visit_start_day_of_week  , SUM(ga_sessions.totals.pageviews) as pageviews
+      FROM filtered_base  AS ga_sessions LEFT JOIN user_label ON ga_sessions.clientId = user_label.clientId  @{QUERY_FILTER} GROUP BY 1,2)) WHERE row_number = 1)),
     -- finding the most common traffic source for the user
       unique_traffic_source AS(
-      (SELECT   ga_sessions_source, fullVisitorId FROM (SELECT ROW_NUMBER () OVER(PARTITION BY fullVisitorId ORDER BY   pageviews) as row_number, fullVisitorId, ga_sessions_source
-      FROM (SELECT ga_sessions.fullVisitorId as fullvisitorid, ga_sessions.trafficsource.medium  AS ga_sessions_source, SUM(ga_sessions.totals.pageviews) as pageviews
-      FROM filtered_base  AS ga_sessions LEFT JOIN user_label ON ga_sessions.fullvisitorid = user_label.fullvisitorid  @{QUERY_FILTER}  GROUP BY 1,2)) WHERE row_number = 1)),
+      (SELECT   ga_sessions_source, clientId FROM (SELECT ROW_NUMBER () OVER(PARTITION BY clientId ORDER BY   pageviews) as row_number, clientId, ga_sessions_source
+      FROM (SELECT ga_sessions.clientId as clientId, ga_sessions.trafficsource.medium  AS ga_sessions_source, SUM(ga_sessions.totals.pageviews) as pageviews
+      FROM filtered_base  AS ga_sessions LEFT JOIN user_label ON ga_sessions.clientId = user_label.clientId  @{QUERY_FILTER}  GROUP BY 1,2)) WHERE row_number = 1)),
 
-      agg_metrics AS (  SELECT  ga_sessions.fullvisitorid, count(distinct visitId) as total_sessions,
+      agg_metrics AS (  SELECT  ga_sessions.clientId, count(distinct visitId) as total_sessions,
         sum(totals.pageviews) as pageviews,
         count(totals.bounces)/count(distinct VisitID) as bounce_rate,
         sum(totals.pageviews) / count(distinct VisitID) as avg_session_depth,
@@ -47,19 +47,19 @@ view: user_facts {
     sum(case when trafficSource.medium = 'cpm' then 1 else 0 end) as visits_traffic_source_cpm,
     sum(case when trafficSource.medium = 'affiliate' then 1 else 0 end) as visits_traffic_source_affiliate,
     sum(case when trafficSource.medium = 'referral' then 1 else 0 end) as visits_traffic_source_referral
-        FROM filtered_base  AS ga_sessions LEFT JOIN user_label ON ga_sessions.fullvisitorid = user_label.fullvisitorid
+        FROM filtered_base  AS ga_sessions LEFT JOIN user_label ON ga_sessions.clientId = user_label.clientId
         @{QUERY_FILTER}   GROUP BY 1 )
 
 
-      SELECT user_label.fullvisitorid, label,ga_sessions_visit_start_hour_of_day, metro, ga_sessions_visit_start_day_of_week, ga_sessions_source,
+      SELECT user_label.clientId, label,ga_sessions_visit_start_hour_of_day, metro, ga_sessions_visit_start_day_of_week, ga_sessions_source,
       total_sessions, pageviews, bounce_rate, avg_session_depth, visits_traffic_source_none, visits_traffic_source_organic, visits_traffic_source_cpc,  visits_traffic_source_cpm, visits_traffic_source_affiliate,
       visits_traffic_source_referral, distinct_dmas, mobile, chrome, safari, browser_other
       FROM user_label
-      LEFT JOIN unique_hour_of_day ON user_label.fullvisitorid = unique_hour_of_day.fullvisitorid
-      LEFT JOIN unique_dma ON user_label.fullvisitorid = unique_dma.fullvisitorid
-      LEFT JOIN unique_day_of_week ON user_label.fullvisitorid = unique_day_of_week.fullvisitorid
-      LEFT JOIN unique_traffic_source ON unique_traffic_source.fullvisitorid = user_label.fullvisitorid
-      LEFT JOIN agg_metrics ON agg_metrics.fullvisitorid = user_label.fullvisitorid
+      LEFT JOIN unique_hour_of_day ON user_label.clientId = unique_hour_of_day.clientId
+      LEFT JOIN unique_dma ON user_label.clientId = unique_dma.clientId
+      LEFT JOIN unique_day_of_week ON user_label.clientId = unique_day_of_week.clientId
+      LEFT JOIN unique_traffic_source ON unique_traffic_source.clientId = user_label.clientId
+      LEFT JOIN agg_metrics ON agg_metrics.clientId = user_label.clientId
        ;;
       persist_for: "24 hours"
   }
@@ -84,9 +84,9 @@ view: user_facts {
   #   sql: TIMESTAMP(PARSE_DATE('%Y%m%d', REGEXP_EXTRACT(_TABLE_SUFFIX,r'\d\d\d\d\d\d\d\d')))  ;;
   # }
 
-  dimension: fullvisitorid {
+  dimension: clientId {
     type: string
-    sql: ${TABLE}.fullvisitorid ;;
+    sql: ${TABLE}.clientId ;;
   }
 
   dimension: label {
@@ -192,7 +192,7 @@ view: user_facts {
 
   set: detail {
     fields: [
-      fullvisitorid,
+      clientId,
       label,
       ga_sessions_visit_start_hour_of_day,
       metro,
@@ -205,4 +205,4 @@ view: user_facts {
       avg_session_depth
     ]
   }
-}
+}
diff --git a/BQML/user_label.view.lkml b/BQML/user_label.view.lkml
diff --git a/Google_Analytics/ga_sessions.view.lkml b/Google_Analytics/ga_sessions.view.lkml
@@ -33,6 +33,7 @@ view: ga_sessions {
     group_label: "ID"
     description: "The unique visitor ID (also known as client ID)."
     sql: ${TABLE}.fullVisitorId ;;
+    hidden: yes
   }
 
   dimension: user_id {
@@ -49,6 +50,11 @@ view: ga_sessions {
     sql: ${TABLE}.visitId ;;
   }
 
+  dimension: client_id {
+    type: string
+    sql: ${TABLE}.clientId ;;
+  }
+
   dimension: visitor_id {
     hidden: yes
     label: "User ID"

diff --git a/marketing.model.lkml b/marketing.model.lkml
@@ -103,11 +103,14 @@ explore: ga_sessions {
 #                   {% endif %};;
 #   }
 
-#   join: future_purchase_prediction {
-#     type: left_outer
-#     sql_on: ${future_purchase_prediction.fullVisitorId} = ${ga_sessions.full_visitor_id} ;;
-#     relationship: one_to_many
-#   }
+}
+
+explore: future_input {
+  join: future_purchase_prediction {
+    type: left_outer
+    sql_on: ${future_purchase_prediction.clientId} = ${future_input.clientId} ;;
+    relationship: one_to_many
+  }
 }