BD-4004 New segment calculation page & additional segmentation edits

[rachel-feinberg]

Why are you making this change? (required)

Add information to historical segment membership docs, including reasons for significant changes. The information comes from a SF KB article listed in the Jira ticket below.

Related PRs, issues, or features (optional)

BD-4004

Feature release date (optional)

• N/A

Contributor checklist

• I confirm that my PR meets the following:
  • I signed the Contribution License Agreement (CLA).
  • My style and voice follows the Braze Style Guide.
  • My content contains correct spelling and grammar.
  • All links are working correctly.
  • If I renamed or moved a file or directory, I set up URL redirects for each file.
  • If I updated or replaced an image, I did not remove the original image file from the repository. (For more information, see Updating an image).
  • If my PR is related to a paid SKU, third party, SMS, AI, or privacy, I have received written approval from Braze Legal.

Submitting for review

If your PR meets the above requirements, select Ready for review, then add a reviewer:

• Non-Braze contributors: Tag @braze-inc/docs-team in a comment below.
• Braze employees: Have any relevant subject matter experts (like engineers or product managers) review your work first, then add the tech writer assigned to your specific vertical. If you're not sure which tech writer to add, you can add braze-inc/docs-team instead.

Thanks for contributing! We look forward to reading your work.

[bclarke-braze]

Just to clarify that we are not just showing the number of users that match within that bucket range.

Suggested change

The historical segment membership count is an estimate, similar to how the segment size is an estimate before you select **Calculate Exact Statistics**. Braze estimates the membership count by querying users in a random bucket range daily at 4 am in your company time zone. This means that on one day, the membership count could be based on users with a random bucket number of 111–120, and on another day, users with a random bucket number of 8,452–8,455. Therefore, the graph might show slight fluctuations on each date due to different amounts of users landing within the random bucket ranges.

The historical segment membership count is an estimate, similar to how the segment size is an estimate before you select **Calculate Exact Statistics**. Braze estimates the membership count by querying users in a random bucket range daily at 4 am in your company time zone and extrapolated to estimate the total count. This means that on one day, the membership count could be based on users with a random bucket number of 111–120, and on another day, users with a random bucket number of 8,452–8,455. Therefore, the graph might show slight fluctuations on each date due to different amounts of users landing within the random bucket ranges.

[wendy-p]

Hey @bclarke-braze could you clarify how this works?

[bclarke-braze]

Yeah, so we take the set of criteria and a random set of bucket numbers as stated above and then we get X amount of users back. So for example, we may check bucket numbers 0-99 and get back 4 users. Because there are 10K buckets, we then essentially multiply 4 by 100 because we were only looking at 1/100 of the user base. So, we'd say the count of users is 400.

Then, based on all the variables, including app group size, we determine a margin of error. The smaller the percentage of users who match in a bucket range and the smaller the bucket range, the larger the margin of error.

The range of bucket numbers we choose is based on the size of the app group. The larger the app group, the fewer buckets we search. We run the query on roughly 50K users, and as stated above, the bucket range may change each day. Unless the amount of users changes drastically, the same amount of buckets would be used each day. So, technically it's not likely to run the query on 10 buckets one day and 14 the next

[wendy-p]

Thanks Brian!!