feat(web-analytics): Fix source query views and uniques - remove blen…

…ded source (#17961)

Fix source query views and uniques

posthog/hogql_queries/web_analytics/ctes.py

@@ -9,35 +9,9 @@
     max(events.timestamp) AS max_timestamp,
     dateDiff('second', min_timestamp, max_timestamp) AS duration_s,
 
-    argMin(events.properties.`$referrer`, events.timestamp) AS earliest_referrer,
-    argMin(events.properties.`$pathname`, events.timestamp) AS entry_pathname,
-    argMax(events.properties.`$pathname`, events.timestamp ) AS exit_pathname,
-    argMax(events.properties.utm_source, events.timestamp) AS earliest_utm_source,
-
-    if(domain(earliest_referrer) = '', earliest_referrer, domain(earliest_referrer)) AS referrer_domain,
-    multiIf(
-        earliest_utm_source IS NOT NULL, earliest_utm_source,
-        -- This will need to be an approach that scales better
-        referrer_domain == 'app.posthog.com', 'posthog',
-        referrer_domain == 'eu.posthog.com', 'posthog',
-        referrer_domain == 'posthog.com', 'posthog',
-        referrer_domain == 'www.google.com', 'google',
-        referrer_domain == 'www.google.co.uk', 'google',
-        referrer_domain == 'www.google.com.hk', 'google',
-        referrer_domain == 'www.google.de', 'google',
-        referrer_domain == 't.co', 'twitter',
-        referrer_domain == 'github.com', 'github',
-        referrer_domain == 'duckduckgo.com', 'duckduckgo',
-        referrer_domain == 'www.bing.com', 'bing',
-        referrer_domain == 'bing.com', 'bing',
-        referrer_domain == 'yandex.ru', 'yandex',
-        referrer_domain == 'quora.com', 'quora',
-        referrer_domain == 'www.quora.com', 'quora',
-        referrer_domain == 'linkedin.com', 'linkedin',
-        referrer_domain == 'www.linkedin.com', 'linkedin',
-        startsWith(referrer_domain, 'http://localhost:'), 'localhost',
-        referrer_domain
-    ) AS blended_source,
+    any(events.properties.$initial_referring_domain) AS $initial_referring_domain,
+    any(events.properties.$set_once.$initial_pathname) AS $initial_pathname,
+    any(events.properties.$set_once.$initial_utm_source) AS $initial_utm_source,
 
     countIf(events.event == '$pageview') AS num_pageviews,
     countIf(events.event == '$autocapture') AS num_autocaptures,
@@ -57,6 +31,19 @@
     ({session_having})
     """
 
+SOURCE_CTE = """
+SELECT
+    events.properties.$set_once.$initial_utm_source AS $initial_utm_source,
+    count() as total_pageviews,
+    uniq(events.person_id) as unique_visitors
+FROM
+    events
+WHERE
+    (event = '$pageview')
+    AND ({source_where})
+    GROUP BY $initial_utm_source
+"""
+
 PATHNAME_CTE = """
 SELECT
     events.properties.`$pathname` AS $pathname,

posthog/hogql_queries/web_analytics/top_pages.py

@@ -42,15 +42,15 @@ def to_query(self) -> ast.SelectQuery | ast.SelectUnionQuery:
 LEFT OUTER JOIN
     (
         SELECT
-            session.entry_pathname,
+            session.$initial_pathname,
             avg(session.is_bounce) as bounce_rate
         FROM
             {session_query} AS session
         GROUP BY
-            session.entry_pathname
+            session.$initial_pathname
     ) AS bounce_rate
 ON
-    pathname.$pathname = bounce_rate.entry_pathname
+    pathname.$pathname = bounce_rate.$initial_pathname
 LEFT OUTER JOIN
     {pathname_scroll_query} AS scroll_data
 ON

posthog/hogql_queries/web_analytics/top_sources.py

@@ -1,7 +1,7 @@
 from posthog.hogql import ast
 from posthog.hogql.parser import parse_select
 from posthog.hogql.query import execute_hogql_query
-from posthog.hogql_queries.web_analytics.ctes import SESSION_CTE
+from posthog.hogql_queries.web_analytics.ctes import SESSION_CTE, SOURCE_CTE
 from posthog.hogql_queries.web_analytics.web_analytics_query_runner import WebAnalyticsQueryRunner
 from posthog.schema import WebTopSourcesQuery, WebTopSourcesQueryResponse
 
@@ -17,25 +17,39 @@ def to_query(self) -> ast.SelectQuery | ast.SelectUnionQuery:
                 timings=self.timings,
                 placeholders={"session_where": self.session_where(), "session_having": self.session_having()},
             )
+        with self.timings.measure("sources_query"):
+            source_query = parse_select(
+                SOURCE_CTE,
+                timings=self.timings,
+                placeholders={"source_where": self.events_where()},
+            )
         with self.timings.measure("top_sources_query"):
             top_sources_query = parse_select(
                 """
 SELECT
-    blended_source as "Source",
-    count(num_pageviews) as "context.columns.views",
-    count(DISTINCT person_id) as "context.columns.visitors",
-    avg(is_bounce) AS "context.columns.bounce_rate"
+    source_query.$initial_utm_source as "Initial UTM Source",
+    source_query.total_pageviews as "context.columns.views",
+    source_query.unique_visitors as "context.columns.visitors",
+    bounce_rate.bounce_rate AS "context.columns.bounce_rate"
 FROM
-    {session_query}
+    {source_query} AS source_query
+LEFT JOIN  (
+        SELECT
+            session.$initial_utm_source,
+            avg(session.is_bounce) as bounce_rate
+        FROM
+            {session_query} AS session
+        GROUP BY
+            session.$initial_utm_source
+    ) AS bounce_rate
+ON source_query.$initial_utm_source = bounce_rate.$initial_utm_source
 WHERE
-    "Source" IS NOT NULL
-GROUP BY "Source"
-
+    "Initial UTM Source" IS NOT NULL
 ORDER BY "context.columns.views" DESC
 LIMIT 10
                 """,
                 timings=self.timings,
-                placeholders={"session_query": session_query},
+                placeholders={"session_query": session_query, "source_query": source_query},
             )
         return top_sources_query