Stacked Bar Chart Total Incorrect for COUNT(DISTINCT) with Dimension (Regression?)

Data Studio Data Studio Q&A
1

Hi Looker Studio team and community,

I think I’ve noticed a change in behaviour recently regarding how totals are calculated in stacked bar charts using COUNT(DISTINCT) metrics alongside a secondary dimension (e.g., a bucket or category).

Previously, when using a chart like:

  • Dimension: MONTH(date)

  • Breakdown dimension: bucket (e.g., days_live_bucket)

  • Metric: COUNT(DISTINCT cla_id)

The total shown in the tooltip or at the top of the stacked column reflected the true number of distinct cla_ids per month, regardless of how many buckets each cla_id appeared in.

Now, the total is calculated as the sum of COUNT(DISTINCT cla_id) across all buckets, which overcounts when a cla_id appears in more than one bucket in the same month. This also affects the percentages shown per bucket — they’re now relative to an inflated total.

This change results in misleading totals and percentages for any metric involving distinct counts across overlapping categories.

:repeat_button: Was this an intentional change?
:wrench: If so, can we have an option to choose between:

  • COUNT(DISTINCT) across total group (true total)

  • Sum of group-level COUNT(DISTINCT) (as currently works)?

This is important for accurate reporting in many business cases (e.g., deduplicated user counts by cohort/bucket).

Thank you for all your work, and I’d love any clarification or update on this.