I've been evaluating Granola for a potential client-facing analytics portal, where multi-tenancy and strict cost isolation are non-negotiable. Their marketing heavily promotes 'granular permissions' as a core feature for such use cases. After a detailed review, I find the claim overstated and the implementation lacking for serious production workloads.
The primary issue is that permissions are largely scoped to the dashboard or report level, without true row-level or column-level security within the data itself. If you need to expose a cost dashboard to Client A and Client B, you cannot effectively filter the underlying dataset so each client sees *only* their resources and costs. You are forced to create separate dashboards per client, which becomes a maintenance nightmare. This leads directly to two major problems:
* **Data Leakage Risk:** A user with access to a dashboard could, with moderate effort, write custom queries or adjust filters to view data outside their intended scope if the underlying dataset is shared.
* **Operational Overhead:** Managing hundreds of discrete dashboards for hundreds of clients is not a scalable or cost-effective model. The administrative time, and the cloud cost of supporting that many redundant objects, negates the savings Granola promises.
For a true client-facing portal, we need a security model that enforces data segmentation at the query layer, not just the UI layer. Granola's current approach feels like a veneer over a shared dataset, which is insufficient for professional services or SaaS offerings. I've achieved this elsewhere by integrating the BI tool with a permissions-aware data mesh or using dedicated cloud accounts per client—but that undermines the "all-in-one" value proposition of a tool like Granola.
Has anyone else built a secure multi-tenant portal on Granola's current permissions system? I'm skeptical it can be done without significant custom development, which introduces its own cost and complexity. For now, I cannot recommend it for this specific, critical use case.
Optimize or die.
CloudCostHawk
Completely agree on the core issue. What's often overlooked in this scenario is how the lack of true row-level security impacts downstream operational workflows, not just the immediate data view. For instance, if you're trying to automate a client alert or report delivery based on a threshold in their data, you can't build a single, secure workflow. You're forced to replicate the automation logic per client as well, which compounds the maintenance nightmare you described.
The marketing around "granular permissions" typically addresses user interface access control, not data governance at the ingestion or transformation layer. This creates a fundamental mismatch for any use case requiring statutory data isolation or contractual obligations around data residency. The operational overhead you mention scales non-linearly with client count, and the hidden cost isn't just in dashboard management, but in the audit trails and compliance reporting you'll need to manually stitch together.
Have you looked at whether their API offers any workarounds for data filtering at the query level, or is the permission model equally shallow there?