Hey <!everyone>! Our 4.1 release is now live: • Holdouts: This oft requested experiment type lets you measure long-term cumulative impact. • Custom Dashboards: Create custom views with key metrics, graphs, charts, and more • AI: Customizable AI to help you write SQL faster, find similar experiments, enforce hypothesis standards, and more • MCP Updates: Create experiments right from your chatbot of choice • Native Vercel Integration: Unified billing, edge SDKs, and central management • Pre-computed Dimensions: Instant breakdowns with zero query time. • FerretDB support: MongoDB compat without the licensing headaches @brief-sugar-18836 • Sanity CMS integration: Create experiments right from your content editor • AND SO MUCH MORE, Y'ALL Feedback always welcome 🫶 and let us know what you want us to build next 🛠️ https://blog.growthbook.io/growthbook-version-4-1/

Hello! Great update, thank you! I've been wanting to check out the dashboards for a long time, and according to the description everything looks great. However, there's one important feature missing for us: writing custom queries (not metrics) and getting the breakdown by variations right inside Growthbook. Sometimes we need to dig deeper, and then we have to copy the query from GB to our BI, which isn't always convenient.

Thanks for the note @late-ambulance-66508. I'll pass it on to the team!

@strong-mouse-55694 Yep, that's possible but only with a huge query with users variations, copied from experiment. I was thinking about another option, where the experiment assignment query can be inserted as a macro and joined to the tables from which the data needs to be collected. For example:

SELECT date, count()
FROM my_table

vs

SELECT date, ab_group, count()
FROM my_table
INNER JOIN {experiment_users} ON ... = ...
...

Currently, copying the query with user distribution seems too cumbersome.

Just for a little clarification, what's an example of what data you're trying to see in the end?

anything in table/chart format for example, we have onboarding flow with 10+ screens, and with the new release we added 1 more for the test group. then, we found significant downgrade in its conversion rate and must analyze is it just a bad screen or there is a bug in implementation (i.e. continue button doesn't work properly), and we need to deep dive in users funnel and check screen by screen conversions. with growthbook, we can answer on the question "what's the result of our test?", but we need detailed analysis to understand what we did better or worse or clarify the next steps.