Hi Team, is it possible for us to use rate as metr...
# ask-questions
Hi Team, is it possible for us to use rate as metrics, such as signup rate in 14 days? If not, how can I calculate the rate changes in specific time frame using count metric?
It is possible - but you need to decide how you're defining this
14 day sign up rate from what date?
you could define it as a 14 day window from the date of exposure to the experiment, and then you would count each registration
Correct. I want A/B testing to compare the signup rate in a 14 days window. I can do this in Google Cloud with BigQuery as following. But I don't know how to track this variation as metric in GB. SELECT COUNT(DISTINCT user_pseudo_id) AS total_users, COUNTIF(event_name = 'sign_up') AS signups, COUNTIF(event_name = 'sign_up') / COUNT(DISTINCT user_pseudo_id) AS signup_rate FROM
WHERE user_pseudo_id is not null and PARSE_DATE('%Y%m%d', event_date) >= DATE_SUB(CURRENT_DATE(), INTERVAL 14 DAY) AND PARSE_DATE('%Y%m%d', event_date) <= CURRENT_DATE()
I think you're over aggregating the metric
I would try something like:
Copy code
  user_pseudo_id as id, -- or whatever your randomization unit is called
  event_date as timestamp, -- assuming this is in the right format
  user_pseudo_id is not null AND
  event_name = 'sign_up'
Defined as a binomial metric. Then, on the metric, you can set the "Metric Window" to "Conversion Window" and then set that to 14 days.
Got it. Thanks.
(under the behavior page)
I set Metric Window to "Conversion Window" and set 14 days with lookback window, but I found when I render this SQL, the date is not set as I expected. The date is start from 20240119, it seems last 90 days duration
The second question
It seems the denominator is not working as expected.