BigQuery: aggregate metrics by running 7d/14d/30d etc. buckets - google-bigquery

Question

Anyone aware of a short, neat BQ query (#standardsql) to aggregate metrics (sessions / PVs / users etc.) by running 7d/14d/30d etc. buckets For ex.
16th-22nd April: 300K sessions
9th-15th April: 330K sessions
2nd-8th April: 270K sessions
OR, out-of-the box function that converts GA's date field (STRING) to days_since_epoch
I wrote a query but it's very complicated
- manually extract as YYY, MM, DD components with REGEXP_EXTRACT()
- convert to days_since_epoch using UNIX_DATE
- divide by '7' to group each row into weekly observations
- use GROUP BY to aggregate & report
any pointers to simplify this use case will be highly appreciable !
Cheers!

Answer 1

Anyone aware of a short, neat BQ query (#standardsql) to aggregate metrics (sessions / PVs / users etc.) by running 7d/14d/30d etc. buckets
See below 7d example for BigQuery Standard SQL - you can apply this logic to whatever data you have with hopefully light adjustments
#standardSQL
WITH data AS (
SELECT
day, CAST(1000 * RAND() AS INT64) AS events
FROM UNNEST(GENERATE_DATE_ARRAY('2017-01-01', '2017-04-25')) AS day
)
SELECT
FORMAT_DATE('%U', day) as week,
FORMAT_DATE('%Y, %B %d', MIN(day)) AS start,
FORMAT_DATE('%Y, %B %d', MAX(day)) AS finish,
SUM(events) AS events
FROM data
GROUP BY week
ORDER BY week
It produces below output that can be used as a starting point for further tailoring to your desired layout
week start finish events
01 2017, January 01 2017, January 07 3699
02 2017, January 08 2017, January 14 4008
03 2017, January 15 2017, January 21 3726
... ... ... ...
OR, out-of-the box function that converts GA's date field (STRING) to days_since_epoch
To convert STRING expressed date into date of DATE type - use PARSE_DATE as in below example
#standardSQL
SELECT PARSE_DATE('%Y%m%d', '20170425') AS date
result is
date
2017-04-25
Finally, below is example/template for running 7d/14d/30d etc. buckets
#standardSQL
WITH data AS (
SELECT
DAY, CAST(1000 * RAND() AS INT64) AS events
FROM UNNEST(GENERATE_DATE_ARRAY('2017-01-01', '2017-04-25')) AS DAY
)
SELECT
DAY,
SUM(CASE WHEN period = 7 THEN events END) AS days_07,
SUM(CASE WHEN period = 14 THEN events END) AS days_14,
SUM(CASE WHEN period = 30 THEN events END) AS days_30
FROM (
SELECT
dates.day AS DAY,
periods.period AS period,
SUM(events) AS events
FROM data AS activity
CROSS JOIN (SELECT DAY FROM data GROUP BY DAY) AS dates
CROSS JOIN (SELECT period FROM (SELECT 7 AS period UNION ALL
SELECT 14 AS period UNION ALL SELECT 30 AS period)) AS periods
WHERE dates.day >= activity.day
AND CAST(DATE_DIFF(dates.day, activity.day, DAY) / periods.period AS INT64) = 0
GROUP BY 1,2
)
GROUP BY DAY
ORDER BY DAY DESC
with output as below
DAY days_07 days_14 days_30
2017-04-25 2087 4004 9700
2017-04-24 1947 4165 9611
2017-04-23 1666 4066 9599
2017-04-22 2121 4820 10014
2017-04-21 2885 5421 10192