2 个特定 Google Analytics（分析）页面点击之间的 Bigquery 时间

Question

纽比在这里。我试图找出我们网站上的用户从一个页面（获取报价）到另一个页面（申请）所花费的平均时间。在 BigQuery 中使用 Google Analytics 数据。我想按频道划分，但最初我只是想为每个用户获取时间。

我尝试过内连接和左连接，但两者都返回 table*table 的结果数。单独运行时，两个子查询看起来都是正确的（引号返回 925，应用程序返回 117），但是当加入它们时，我得到了重复。

注释掉的位用于最终查询，可以忽略，只包括上下文。

-- SELECT
--   channelGrouping,
--   ROUND(AVG(timeQ2A/60000),2) AS avgMinsQ2A
-- FROM (

   SELECT
    CONCAT(CAST(visitId AS STRING),'_',fullVisitorId) AS sessionId,
    apps.channelGrouping,
    (quote.quoteTime) as quoteTime,
    (CAST(hits.time AS INT64)) AS appTime,
    (CAST(hits.time AS INT64) - (quote.quoteTime)) AS timeQ2A
  FROM
    `project.dataset.ga_sessions_20190923` AS apps,
    UNNEST(hits) AS hits
  INNER JOIN (
    SELECT
     MIN(CAST(hits.time AS INT64)) AS quoteTime,
     CONCAT(CAST(visitId AS STRING),'_',fullVisitorId) AS sessionId     
    FROM
      `project.dataset.ga_sessions_20190923`,
      UNNEST(hits) AS hits
    WHERE
      hits.type = "PAGE"
      AND hits.page.pagePath = "/get-a-quote/"
      AND hits.time > 0 
    GROUP BY
    2) AS quote
  ON
    sessionId = quote.sessionId
  WHERE
    hits.type = "PAGE"
    AND hits.page.pagePath = "/application-complete/"
  GROUP BY
    1,
    2,3,4
ORDER BY 1,2,3
--     )
-- GROUP BY
--   1
-- ORDER BY 
--   2 DESC

我希望只获得 117 行（所有已申请的用户都将在报价表中，因此内部和左侧应该相同），但我得到 108k。

非常感谢任何帮助。

Answer 1

如果我不得不猜测，我会说您在每个会话中有多个点击，其中hits.page.pagePath = "/application-complete/" 为真。您在内部查询中获取时间戳的最小值，而不是外部查询。

如果是我，我会像这样构造我的查询：

-- Get all relevant GA Data
with ga as (
  select
    CONCAT(CAST(visitId AS STRING),'_',fullVisitorId) AS sessionId,
    channelGrouping,
    hits.time as hit_ts,
    hits.page.pagePath
  from `project.dataset.ga_sessions_20190923`
  left join unnest(hits) AS hits
  where hits.type = 'PAGE' and hits.page.pagePath in ('/get-a-quote/','/application-complete/') and hits.time > 0
),
-- Get first instance of quote
first_quote as (
  select
    sessionId,
    channelGrouping,
    min(hit_ts) as quote_ts
  from ga
  where pagePath = '/get-a-quote/'
  group by 1,2
),
-- Get first instance of app-complete
first_app_complete as (
  select
    sessionId,
    channelGrouping,
    min(hit_ts) as app_complete_ts
  from ga
  where pagePath = '/application-complete/'
  group by 1,2
),
-- Join it all together and calc differences
joined as (
  select
    sessionID,
    channelGrouping,
    timestamp_diff(app_complete_ts, quote_ts, second) as ts_diff_seconds
  from first_quote
  inner join first_app_complete using(sessionID,channelGrouping)
)
select * from joined
-- select channelGrouping, count(sessionID) as session_count, avg(ts_diff_seconds) as average_seconds_to_complete from joined group by 1