Context
PR #7215 introduced label-based grouping for Feature Analytics charts: when the API returns labelled evaluation buckets (per-SDK user_agent / client_application_name), the chart stacks bars by label value instead of by environment.
This surfaced a UX question that wasn't resolved in that PR:
The problem
When the chart is in labelled mode, the environment selector at the top still works but its effect is invisible — selecting Production + Staging vs. just Production changes the stack totals silently, with no way to tell which environment contributed to each SDK slice.
Example: a bar showing flagsmith-js-sdk: 500 could be 300 from Production + 200 from Staging, or 500 from Production alone. Deselecting an env re-aggregates the total, but there's no visible breakdown.
What was considered in #7215
- Hide the env selector in labelled mode — simpler, matches the org-level Usage page pattern (which has no env selector in SDK view). Risk: users lose the ability to narrow the dataset to a specific env at all.
- Keep the env selector as a silent data filter — this is what shipped. It still narrows the dataset; users just can't see the breakdown inside each SDK stack.
- Nested grouping (env × SDK) — each env becomes a sub-stack within each SDK bar. Visually complex, potentially noisy with 5 envs × 5 SDKs.
Open questions
- Is per-environment breakdown valuable when labels are present, or is SDK breakdown sufficient at that level of aggregation?
- If both are valuable, what's the right UX? Options:
- Toggle between "Group by Environment" / "Group by SDK" (mode switch)
- Show env names in the tooltip (non-stacked, but informative)
- Nested stack (env × SDK)
- Let the user drill down from SDK → env on click
- Does the answer differ for single-env vs multi-env selection?
Not in scope for #7215
The PR ships the simpler "labelled mode replaces env mode" behaviour. This issue tracks the follow-up UX decision.
References
Context
PR #7215 introduced label-based grouping for Feature Analytics charts: when the API returns labelled evaluation buckets (per-SDK
user_agent/client_application_name), the chart stacks bars by label value instead of by environment.This surfaced a UX question that wasn't resolved in that PR:
The problem
When the chart is in labelled mode, the environment selector at the top still works but its effect is invisible — selecting Production + Staging vs. just Production changes the stack totals silently, with no way to tell which environment contributed to each SDK slice.
Example: a bar showing
flagsmith-js-sdk: 500could be 300 from Production + 200 from Staging, or 500 from Production alone. Deselecting an env re-aggregates the total, but there's no visible breakdown.What was considered in #7215
Open questions
Not in scope for #7215
The PR ships the simpler "labelled mode replaces env mode" behaviour. This issue tracks the follow-up UX decision.
References
sdk_usage_chartsflag) — no env selector at all in that view