Counting google analytics unique events in BigQuery

问题

I have managed to calculate total events by ISOweek but not unique events for a given Google Analytics Event using BigQuery. When checking GA, total_events matches the GA interface on the dot but unique_events are off. Do you know how I can solve this?

The query:

SELECT INTEGER(STRFTIME_UTC_USEC(PARSE_UTC_USEC(date),"%V")) iso8601_week_number,
hits.eventInfo.eventCategory,
hits.eventInfo.eventAction,
COUNT(hits.eventInfo.eventCategory) AS total_events,
EXACT_COUNT_DISTINCT(fullVisitorId) AS unique_events
FROM
    TABLE_DATE_RANGE([XXXXXX.ga_sessions_], TIMESTAMP('2017-05-01'), TIMESTAMP('2017-05-07'))
WHERE
  hits.type = 'EVENT' AND hits.eventInfo.eventCategory = 'BIG_Transaction'
GROUP BY
iso8601_week_number, hits.eventInfo.eventCategory, hits.eventInfo.eventAction

回答1:

The definition of unique events in Google Analytics is:

A count of the number of times an event with the category/action/label value was seen at least once within a session.

In other words, the number of sessions in which a specific event (defined by category, action AND label) was sent. In your query, you count the number of unique visitors that had the event, while you need to count the number of sessions and keep in mind that events with different labels should be counted as different unique events (although we are only interested in category and action).

A possible way to fix your code is:

SELECT 
  INTEGER(STRFTIME_UTC_USEC(PARSE_UTC_USEC(date),"%V")) iso8601_week_number,
  hits.eventInfo.eventCategory,
  hits.eventInfo.eventAction,
  COUNT(hits.eventInfo.eventCategory) AS total_events,
  EXACT_COUNT_DISTINCT(CONCAT(fullVisitorId,'-',string(visitId),'-',date,'-',ifnull(hits.eventInfo.eventLabel,'null'))) AS unique_events    
FROM
  TABLE_DATE_RANGE([XXXXXX.ga_sessions_], TIMESTAMP('2017-05-01'), TIMESTAMP('2017-05-07'))
WHERE
  hits.type = 'EVENT' AND hits.eventInfo.eventCategory = 'BIG_Transaction'
GROUP BY
    iso8601_week_number, hits.eventInfo.eventCategory, hits.eventInfo.eventAction

The results of this query should match with the data in the GA interface.

回答2:

Depending on the scope you need to count(distinct ) different things, but you always need to fulfill these conditions:

unique events refer to the combination of category, action and label
make sure eventAction is not NULL
make sure eventLabel is not NULL
eventCategory is allowed be NULL

I'm using COALESCE() to avoid NULLs

Example Session Scope

SELECT
  SUM( (SELECT COUNT(h.eventInfo.eventCategory) FROM t.hits h) ) events,
  SUM( (SELECT COUNT(DISTINCT 
    CONCAT( h.eventInfo.eventCategory,
      COALESCE(h.eventinfo.eventaction,''),
      COALESCE(h.eventinfo.eventlabel, ''))
      )
    FROM
      t.hits h ) ) uniqueEvents
FROM
  `google.com:analytics-bigquery.LondonCycleHelmet.ga_sessions_20130910` t

Example Hit Scope

SELECT
  h.eventInfo.eventCategory,
  COUNT(1) events,
  -- we need to take sessions into account, so we add fullvisitorid and visitstarttime
  COUNT(DISTINCT CONCAT(fullvisitorid, CAST(visitstarttime AS string), 
    COALESCE(h.eventinfo.eventaction,''), 
    COALESCE(h.eventinfo.eventlabel, ''))) uniqueEvents
FROM
  `google.com:analytics-bigquery.LondonCycleHelmet.ga_sessions_20130910` t,
  t.hits h
WHERE
  h.type='EVENT'
GROUP BY
  1
ORDER BY
  2 DESC

hth!

回答3:

I believe the issue is that you are only counting the number of unique visitors have completed the specified action, while GA defines unique events as "The number of times during a date range that a session contained the specific dimension".

Therefore, I would just change your code to the below:

SELECT INTEGER(STRFTIME_UTC_USEC(PARSE_UTC_USEC(date),"%V")) iso8601_week_number,
hits.eventInfo.eventCategory,
hits.eventInfo.eventAction,
COUNT(hits.eventInfo.eventCategory) AS total_events,
EXACT_COUNT_DISTINCT(CONCAT(fullVisitorId, STRING(visitId))) AS unique_events
FROM
    TABLE_DATE_RANGE([XXXXXX.ga_sessions_], TIMESTAMP('2017-05-01'), TIMESTAMP('2017-05-07'))
WHERE
  hits.type = 'EVENT' AND hits.eventInfo.eventCategory = 'BIG_Transaction'
GROUP BY
iso8601_week_number, hits.eventInfo.eventCategory, hits.eventInfo.eventAction

This should give you the distinct count of sessions that had the given events.

来源：https://stackoverflow.com/questions/44203413/counting-google-analytics-unique-events-in-bigquery

标签

google-analytics

google-bigquery