Frequently asked questions

BI Pixie is a usage-tracking and AI Readiness platform for Power BI. It adds Pixies, invisible tracking pixels native to Power BI, to your reports, and through them it captures engagement: page views, visual clicks, filter selections, and feedback. BI Pixie collects viewer user names only when you turn on Capture user identity, and until then it counts viewers by a masked IP address. It also makes your semantic models and Fabric data agents ready for AI, and proves AI accuracy on your own data. See the Product page for the full capability list.

Native usage metrics show page-level views and a list of viewers. BI Pixie tracks visual-level clicks, filter and slicer selections, bookmark navigation, drill-through, and user feedback, and it covers embedded and publish-to-web reports, which native usage metrics do not cover at all. See what usage metrics leave out.

AI Readiness tells you whether your semantic models and Fabric data agents are ready for Copilot and AI, and what to fix first. BI Pixie assesses each one and drafts the fixes. A data agent is assessed whatever data it is grounded on: semantic models, lakehouses, warehouses, KQL, SQL and mirrored databases. You review every change and choose what gets written. AI Readiness is available in BI Pixie Cloud and the Fabric Workload. See the AI Readiness guide.

Most teams start with BI Pixie Cloud, which is fully managed and live in minutes. If you already work inside Microsoft Fabric, the Fabric Workload gives you the same product natively in the Fabric portal. If you need full data sovereignty or already run in your own Azure or Power Platform environment, self-hosted keeps your telemetry in your tenant, with usage tracking only (no AI Readiness). See the full deployment comparison.

You set up your AI so that BI Pixie can write text for you, check its own findings, and stand in for an AI under test: proposed descriptions, AI Instructions, display folder names, confirmation that two similarly named fields are really duplicates, the business domains and natural phrasing behind benchmark questions, the second opinion on each benchmark answer, and the generic AI agent that a benchmark can test. Scoring never uses it, and neither does the dimension breakdown, the findings and the order they come in, usage grounding, assessment history, or field defaults. AI data schema curation works before your AI is set up and uses it only for the duplicate confirmation named above. A benchmark can still be built and run before your AI is set up, using fixed question templates instead of natural phrasing.

BI Pixie calls your AI when you ask it to draft something, or when a benchmark you have set up runs. Usage therefore follows the work you chose to do, such as the size of the work list you asked BI Pixie to draft, rather than how much your reports are used. BI Pixie sends the part of the semantic model definition the task needs, meaning names, existing descriptions, measure expressions and AI Instructions, and where your reports have Pixies it also sends the report, page and visual titles people open and the values people filter to most. Your AI never receives telemetry event records or end-user identifiers. Privacy Policy Section 3.7 states what is sent and what is never sent.

You can bring your own AI, in which case the usage lands on your own account under your own limits and controls: Azure AI Foundry runs in your own Azure tenant, and OpenAI and Anthropic each take an API key. Accounts with a designated Fabric workspace can instead use the BI Pixie data agent, which runs on your own Fabric capacity with no API key to manage. If you would like per-operation estimates for planning, ask us and we will share what we measure. See AI Setup for setup.

Only where you approve a specific change. Assessing is read-only: BI Pixie reads the definition through the Fabric API using your own identity, so it can never reach a semantic model you cannot, and it writes nothing. When you optimize, each description, instruction, or folder arrives as a draft you can edit or discard, and nothing reaches the semantic model until you apply it. Writes go through your own Fabric permissions, so you need Contributor access or higher on the workspace.

No. BI Pixie Cloud is a hosted service: you sign in, connect your Power BI workspaces, add Pixies to your reports, and install the BI Pixie Dashboard into a workspace of your own. A Power BI Pro workspace is enough, and it supports up to eight dashboard refreshes a day. A Fabric capacity matters in four places: running BI Pixie as a Fabric Workload, creating a Fabric data agent, benchmarking through Copilot or through your own data agent, and moving the telemetry pipeline to a Spark ETL once your volumes outgrow an import refresh.

Not for assessing or optimizing. Both read the semantic model definition through the Fabric API with your own identity and run in BI Pixie’s service rather than on your capacity, so they work on Power BI Pro and Premium Per User workspaces. A benchmark depends on which AI you test: Copilot and your own Fabric data agent both answer on your capacity and need one, while a generic AI agent answers through your AI and needs none. AI to Test carries the full table.

BI Pixie can run natively in Microsoft Fabric, either as the BI Pixie Workload added from the Workload Hub, or installed into a Fabric workspace from BI Pixie Cloud. Either way a Fabric notebook processes your engagement data on your own capacity and the dashboard reads the result through Direct Lake, so there is no import refresh to time out. That matters most for a large Power BI estate with heavy report usage, where the volume of engagement events outgrows an import refresh, and your telemetry stays in your own tenant governed by your own policies. If you do not use Fabric but already run something like Azure Databricks, the same ETL can run there instead: contact us and we will provide the scripts. See Dashboard refresh for how the two pipelines compare.

Two ways forward. Ask your Fabric administrator to add BI Pixie from the Fabric Workload Hub, which is a tenant-level action, after which anyone in the tenant can create a BI Pixie item from a workspace’s New menu. Or start in BI Pixie Cloud and install BI Pixie into a Fabric workspace you already have, from the Overview page. That path does not need the workload approved, and you can move to the workload later.

BI Pixie has three plans: Standard, Pro, and Enterprise, on top of the Free plan every account starts on. Plans differ in tracked items, benchmark runs, report viewers, and support. See Pricing for current details, or the Free Plan section at the top of this page for what is included free.

Yes. The Free plan is a plan in its own right, not a trial with a countdown. It gives you an AI Readiness assessment for up to 500 semantic models and data agents, one tracked report, one tracked semantic model or data agent, and one AI benchmark run on your tracked semantic model. That is enough to assess your mission-critical semantic models and data agents, measure how one report is used, and prove the result on one semantic model before you decide anything. See the Free Plan section at the top of this page for the full list.

No. One BI Pixie subscription, through the Azure Marketplace or the BI Pixie portal, covers BI Pixie Cloud and the Fabric Workload. Self-hosted deployments are sold as part of the Enterprise plan.

Yes. Upgrades take effect immediately. Downgrades take effect at the end of your current billing cycle. See the Pricing FAQ for proration details.

It depends on your deployment. BI Pixie Cloud stores telemetry in BI Pixie's hosted infrastructure, with Bring Your Own Storage available. The Fabric Workload stores telemetry in BI Pixie's Azure by default, with an option to keep it in your own OneLake for data residency. Self-hosted keeps every event in your own Azure tenant or Power Platform environment. See each deployment's data residency docs for the details.

BI Pixie collects viewer user names only when you turn on Capture user identity in Tracking Setup, which is off by default. Until then, BI Pixie counts viewers by a masked IP address, so it measures engagement rather than who generated it. When you turn it on, you choose on the Account page whether BI Pixie keeps each viewer's email address, a masked ID or nothing.

Not for BI Pixie Cloud or the Fabric Workload; both are fully hosted. For self-hosted, you deploy a small set of components into your own Azure tenant or Power Platform environment: a few Azure Functions, storage, and a small Windows app, BI Pixie Desktop (bi-pixie.exe), that adds Pixies to your reports. See the docs for your deployment for setup steps.

This page covers general questions. For step-by-step setup, tracking configuration, and troubleshooting on your specific deployment, see the Cloud FAQ, the Fabric Workload FAQ, or the self-hosted guide.

Visit the Support page, or contact us directly. Pro and Enterprise plans include email support.