Hi everyone, I’m hoping for some advice. I’m a bit out of my depth here, so apologies if this is a basic question.
We recently upgraded our BI platform (not naming names, but it’s a major cloud one) specifically for its advertised ability to handle 50+ concurrent users for self-serve reporting. Our marketing team is about 25 people, but when we run our weekly performance deep dives, maybe 15 of us are in the tool at once building reports. That’s when things start to slow down dramatically, sometimes to the point of timeouts. The vendor’s documentation says we should be well within the limits.
I’m nervous because I pushed for this tool based on those specs, and now my team is getting frustrated. We’re connecting it to our data warehouse, and our dashboards use live connections, not extracts. Could that be the issue? Or is there something in the configuration we might have missed? Our use case is mainly for the marketing team to analyze campaign performance and map it to lead stages in our CRM.
Has anyone else run into this gap between advertised concurrency and real-world performance? Were you able to fix it through settings, or did you have to change your data architecture? Any guidance would be so appreciated.
🙏 jane
Oh, I've been there, and the frustration is real. It sounds like you're right on the edge of the vendor's happy path, but the reality is that "concurrent users" is rarely a simple headcount metric.
>dashboards use live connections, not extracts
This is very likely a huge part of the problem. When 15 people are all running queries directly against your warehouse at once, you're hitting two potential bottlenecks: the BI tool's own server capacity and your data warehouse's concurrent query limits. The vendor's concurrency number often assumes a mix of light browsing and maybe a few builders, not 15 people running complex, live queries simultaneously.
My first step would be to check the query logs in your data warehouse (BigQuery, Snowflake, Redshift etc.) during one of those slowdowns. You might be getting queued or throttled there before the BI tool even breaks a sweat. For a temporary fix, could you schedule a high-use report to refresh as an extract right before the meeting? It's not ideal for self-serve, but it could confirm the architecture is the choke point. Also, peek at your user management settings - sometimes there are "viewer" vs "editor" tiers with different resource allocations.
Stay connected