Beyond the Connector: Enterprise OneStream Analytics in Power BI and Copilot
The OneStream Power BI connector gets your OneStream data into Power BI. A governed semantic model is what turns that data into reports and AI answers your whole finance team can trust. In this on-demand webinar, Ryan Connors, Group Product Director at OneStream, walks through what the certified connector does today and what’s coming next. Marla Nelson, CTO at QuickLaunch Analytics, then shows the five steps that take OneStream cube data to an enterprise semantic model, with a live demo. David Kettinger, Director of Marketing at QuickLaunch Analytics, hosts.
Ryan and Marla first gave a version of this session to a full room at OneStream Splash in May. This one goes further into AI.
Chapters
- 01:24 Welcome and housekeeping
- 03:00 Agenda and an introduction to QuickLaunch Analytics
- 04:42 Meet the speakers
- 05:50 Why OneStream and QuickLaunch work together
- 07:33 Inside the OneStream Power BI connector
- 10:31 The four Analytics APIs
- 14:02 Field tip: keeping security once you publish to Power BI Service
- 15:23 Field tip: working past the 2 million record limit
- 17:12 What’s coming for the connector
- 19:15 Five steps from the connector to an enterprise semantic model
- 22:32 Step 1: future-proof the architecture
- 24:52 Step 2: turn cube data into a star schema
- 27:35 Step 3: standardize metrics and financial intelligence
- 29:11 Step 4: reports, paginated reports and conversational analytics
- 30:24 Step 5: governance and scale
- 31:13 Demo: product and market analysis
- 33:31 Demo: KPI report
- 34:42 Demo: payroll and workforce planning with blended data
- 36:47 Demo: hierarchies and drill-through to ERP detail
- 40:17 Demo: paginated financial statements
- 41:57 Demo: conversational analytics and AI financial briefings
- 45:39 Demo: semantic model documentation report
- 46:36 Closing points
- 48:02 Live Q&A
What does the OneStream Power BI connector do?
The connector is OneStream’s Microsoft-certified way to pull OneStream data into Power BI Desktop or Microsoft Fabric. You find it in the Microsoft connector list, sign in with your OneStream credentials, and the connector applies your OneStream security, so you only pull the data you’re allowed to see. From there you publish to Power BI Service and schedule refreshes. Ryan said most customers build their datasets, check them, then set them to refresh daily, several times a day or monthly.
Along with the connector, OneStream released the Analytics API. It has four calls, two for data and two for metadata.
- Get Custom Data Adapter. Ryan called this the special sauce. It pulls any data adapter you build in OneStream: a cube view, a CubeViewMD, a SQL query against staging data, a method query or a business rule. Because a cube view comes across exactly as it appears in OneStream, you get values that OneStream calculates on the fly, such as consolidated totals and cube view math.
- Get Cube Data. Think of it as a data management extract. It pulls stored data at base level or any hierarchy level once you supply the dimensionality, from entity and account down to the user-defined dimensions.
- Get Dimensions. Returns your hierarchies pre-leveled, so Power BI can rebuild them quickly instead of you flattening them by hand.
- Get Member Properties. Adds member details used in reporting, such as text fields, descriptions and currency.
Two tips from the field
Security behaves differently once you publish. In Power BI Desktop, the connector enforces your OneStream access. In Power BI Service the live connection is gone, so the published report doesn’t carry that security on its own. Ryan described the pattern customers use: pull OneStream’s security tables with a method query, load them into Power BI, and point row-level security at those tables. You don’t maintain access by hand, and you can still grant access to people who aren’t OneStream users.
Each API call returns up to 2 million records. OneStream stores data by data unit (entity, scenario and time), so the fix is to build a list in Power BI Desktop, for example of entities, and loop the same call over that list. Power BI stitches the results back together, the refresh can still run on a schedule, and smaller calls put less load on OneStream. OneStream has a video on this pattern for customers.
What’s coming for the connector
Ryan shared three enhancements OneStream is targeting for the first half of 2027. He was clear that this is forward-looking and could change, and that the release also needs Microsoft recertification.
- Partitioning inside the API, so you no longer build the entity list yourself.
- Parquet as a transfer format, which is more compressed, moves faster and puts less stress on OneStream.
- Flexible member properties, so you pick the properties you want instead of taking a fixed set.
He added that existing pulls will keep working as they do today.
Five steps from the connector to an enterprise semantic model
Marla’s point was simple. The connector is step one. If every report gets its own cube view or extract, you end up with many small semantic models that each define revenue their own way. The goal is one governed semantic model that every report and every AI tool reads from.
Step 1: Future-proof the architecture
Plan how you’ll extract and structure the data before you build, so you aren’t rebuilding six months from now. QuickLaunch installs a solution workspace and ready-built data adapters in your OneStream environment, so refreshes can start right away. A dimension mapping table lives in OneStream, which keeps OneStream as the system of record and tells Power BI which accounts make up revenue and your other KPIs. For volume, QuickLaunch partitions the fact table, so a monthly refresh reloads the dimension tables and the current partition rather than every year of history.
Step 2: Turn cube data into a star schema
OneStream cube data doesn’t look like a relational database, and rebuilding it as a star schema in Power BI takes real semantic modeling skill. Marla covered the pieces. A calendar dimension makes year-to-date and prior-year comparisons work, including 4-4-5 and offset fiscal calendars, and the connector doesn’t include one. The fact table holds actuals, budget and forecast at the lowest grain OneStream has, as monthly (or weekly) activity plus ending balances. Entity, account, scenario and the user-defined dimensions sit around it, with main and alternate hierarchies rebuilt to match OneStream. Sort order is the part teams miss. Power BI sorts members alphabetically by default, and QuickLaunch extends the connector so an income statement reads revenue, cost of sales, gross profit in OneStream’s order.
Step 3: Standardize metrics and financial intelligence
This is where build-as-you-go projects stall. One analyst’s revenue doesn’t match another’s. Standard measures fix that, along with sign handling, so a revenue variance reads favorable and an expense variance reads the right way. OneStream also stores data at every entity level, so a careless measure can add parent and child entities together. QuickLaunch protects its measures against summing across entities and builds in side-by-side scenario comparisons with variances and percentage differences. The OneStream Application Pack includes 900+ calculations.
Step 4: Reports and AI
People still need interactive dashboards with KPIs, drill-through and mobile, and they still need pixel-perfect paginated reports for formal income statements and balance sheets. Conversational analytics is now part of the mix too, through Microsoft 365 Copilot, Fabric data agents and Claude. Because each one reads the same governed semantic model, your reports, your analysts and your AI all work from the same trusted definitions.
Step 5: Governance and scale
Mirror OneStream’s security in the semantic model, so people see only the data they’re allowed to see, and extend it to people outside OneStream when you need to. Set a refresh strategy with incremental refresh, a dev, test and production promotion path, and usage monitoring and lineage that satisfy your controls.
What the demo showed
Marla worked through a set of reports built on the QuickLaunch semantic model for OneStream.
- A product and market analysis (using UD2 for product and UD3 for market) that drilled from a dip in gross profit percent to the public sector market and then to the product line behind it, switching between year-to-date, quarter-to-date and month-to-date.
- A KPI report with ratios such as current ratio, quick ratio and AR turnover, each with a rolling 12-month trend of actual against plan.
- Payroll actual against plan using statistical data loaded in OneStream, such as headcount, and a workforce projection that blends OneStream data with a headcount forecast kept in an Excel file in SharePoint.
- A drill from a OneStream trial balance variance (a rent account at $115,000 against a $67,000 plan, in demo data) across to general ledger journal entries and the AP voucher in an ERP semantic model. That drill across semantic models works when your OneStream data sits beside your ERP data in the same foundation.
- A paginated income statement with the same drill-through.
- A plain-English question about operating expenses against budget that came back as a waterfall chart, and an AI financial briefing built as a Fabric data app. The briefing reads the semantic model, summarizes results, flags favorable and unfavorable lines, and includes a trust panel so readers can see where the figures come from.
- A documentation report showing the scope of the semantic model, its partitioning by scenario, fiscal period and currency, and its security roles.
Where to start
Use the certified connector. It’s the supported, validated path from OneStream to Power BI. Then define the hierarchies and the key metrics once, in one governed semantic model, so every report works from the same set of numbers. Carry OneStream’s security across, and extend it on purpose to people outside OneStream. And before you point AI at OneStream data, get that semantic model right, because the AI reads what the semantic model tells it.
Free eBook
Top 8 OneStream Analytics Challenges and How to Solve in Power BI
Frequently asked questions
Is there a OneStream connector for Tableau?
Yes. OneStream sells its connectors as a BI Connectors bundle that covers Power BI and Tableau, and the Tableau version was released in 2026. It isn’t in the Tableau store. You load a solution into OneStream, and it pushes data to Tableau as a Hyper extract file. Ryan said the functionality matches the Power BI connector.
Should you extract YTD or periodic Income Statement data into Power BI?
Both. QuickLaunch brings in ending balances and monthly activity in the same pull. Ending balances are inception-to-date for balance sheet accounts and year-to-date for income statement accounts. Monthly activity covers both. For income statement reporting, the semantic model typically sums monthly activity to calculate year-to-date, quarter-to-date and rolling 12-month views, so any time comparison works from one source.
Can Power BI write data back to OneStream through the connector?
No. The OneStream Power BI connector and its Analytics API are read-only and move data from OneStream to Power BI. If you need to send values back, such as a reason captured during analysis, you’d build your own process using OneStream’s data integration tools to load the data into OneStream.
What does it take to set up the OneStream Power BI connector, and who owns it?
Once you buy the connector, OneStream’s cloud team turns on access to the Analytics API and you can start pulling data. Ryan said you shouldn’t need IT for setup. The projects that go best put a OneStream expert and a Power BI expert in the same room, so the person who knows the data and the person who knows the semantic model agree on what to bring over.
What happens to Power BI reports when finance changes a OneStream hierarchy or adds an account?
New members and changes to sort order come through on the next refresh of the dimension tables, which QuickLaunch recommends refreshing in full every time, even during an incremental refresh. A brand-new hierarchy level needs a change to the semantic model, with or without QuickLaunch. QuickLaunch builds hierarchies from the top down and relates them at the lowest member level, so adding a level typically doesn’t break existing reports.
What do you need in place before pointing AI tools like Copilot or Claude at OneStream data?
A strong semantic layer. Marla’s view is that AI answers depend on the business definitions underneath, such as what each region means and how each metric is calculated. With a governed semantic model in place, Microsoft 365 Copilot and Claude answer from the same trusted definitions as your reports.
