feat(insights): Warn about WAU/MAU in total value trends #21067
Conversation
| WorldMap = 'WorldMap', | ||
| BoldNumber = 'BoldNumber', | ||
| } | ||
| export enum ChartDisplayCategory { |
There was a problem hiding this comment.
Added ChartDisplayCategory as an abstraction over the specific display types
I also wanted to use it in the backend, the way we reuse ChartDisplayType thanks to schema.{ts,json,py}, but unfortunately ts-json-schema-generator doesn't support re-exporting, so we can't do that (unless we start using ChartDisplayCategory in a type defined in schema.ts itself). I put out a fix for the library (super interesting opportunity to learn more about TS syntax internals!), but in the meantime it's fine for this to be frontend-only
There was a problem hiding this comment.
ChartDisplayCategory would specifically be super useful for caching
|
Size Change: +261 B (0%) Total Size: 824 kB ℹ️ View Unchanged
|
|
Might it be better if we hide these two items for total value insights? Or other way around: When would I use the combination of these two? Right now I find the warning a bit strongly worded and would go for something like "in most cases you want to use unique users, only for ... it makes sense to use wau/mau". |
|
Hmm, hiding is the trickier option, because people switch between time series and total value all the time (like the user in the ticket), so the solution wouldn't be complete without accounting for that workflow. So either: Option 1: We only hide MAU/WAU on total insights if not already used – but then we still need the warning to inform about the problem Option 2: We switch from MAU/WAU to "Unique users" for the user – only the results would still be unexpected, because there's the same problem: what date range should MAU/WAU mean in a total value context? I think the only solution that isn't surprising would be a modal on display type change to "Hey, what date range do you want for Unique users specifically?" But this also breaks down when there are multiple series where not all are MAU/WAU. Hence leaning towards just the warning for now. I think what'd be amazing (once HogQL trends are the only trends) would be a series-level rolling average option to replace both WAU/MAU and the current "smoothing" feature. Basically this: #11716 |
| In total value insights, it's usually not clear what date range "{definition.name}" refers | ||
| to. For full clarity, we recommend using "Unique users" here instead |
There was a problem hiding this comment.
I reworded this a bit to hopefully suggest "Unique users" more gently @thmsobrmlr The tricky bit is that it's really hard to understate how confusing MAU/WAU is in this situation 😅
Yup, that makes sense. One thought I had: We could have a general warning area of some form for insights, since people might easily miss the warning in the dropdown when switching between time series and total value insights. We could even add an
Oh yeah, that looks good. Didn't know about this suggestion at all. |
frontend/src/scenes/insights/filters/ActionFilter/ActionFilterRow/ActionFilterRow.tsx
Outdated
Show resolved
Hide resolved
…Row/ActionFilterRow.tsx Co-authored-by: Thomas Obermüller <[email protected]>
📸 UI snapshots have been updated1 snapshot changes in total. 0 added, 1 modified, 0 deleted:
Triggered by this commit. |
📸 UI snapshots have been updated1 snapshot changes in total. 0 added, 1 modified, 0 deleted:
Triggered by this commit. |
📸 UI snapshots have been updated1 snapshot changes in total. 0 added, 1 modified, 0 deleted:
Triggered by this commit. |
📸 UI snapshots have been updated1 snapshot changes in total. 0 added, 1 modified, 0 deleted:
Triggered by this commit. |
Problem
ZEN-11743 showcases how confusing and prone to errors the combination "Weekly/Monthly active users" math with total value trends is. For example, when you have a date range that's "Last 90 days", what does "Monthly active users" mean? There really is no intuitively correct answer that I can think of, so we should at the very least discourage this combination.
Changes
Added a warning for WAU/MAU series on total value trends:
How did you test this code?
Added an E2E test.