
What Is an OData Feed in Power BI?
Carlos GarciaPublished Somewhere in every Power BI tutorial, somebody tells you to "just connect to the OData feed," as though you arrived already knowing what one is. The menu item is real enough — it sits in Get data under the name OData Feed — but the documentation around it tends to assume you already know what an OData service is and what you are supposed to type into the box.
This guide works from Microsoft's own pages: the Power Query OData Feed connector reference, Microsoft's guidance for reading Azure DevOps Analytics into Power BI, and Microsoft's performance notes for Business Central APIs. Where Microsoft does not state something, this guide says so rather than quietly filling the gap. There is one important thing it never states, and that absence is the most useful fact in this article.
The short version: an OData feed is a web address that returns data in a standard, predictable shape, and Power BI can push your filtering back to the server instead of dragging everything down first. The long version is where the money is.
What Is an OData Feed in Power BI?
An OData feed is a live data endpoint published over HTTP using OData, a standard way of exposing tables and their relationships as URLs. Instead of handing Power BI a file, you hand it an address. Power BI reads the service description at that address, discovers which tables exist, and lets you choose the ones you want.
In Power BI, that happens through the connector Microsoft calls OData Feed. Its reference page lists support in Excel, Power BI (semantic models) and Power Query Online, with a release state of General availability. This is a mature, first-party connector rather than a preview feature, and the page carries no deprecation, retirement or replacement notice of any kind.
What separates an OData feed from a plain web request is that it is queryable. Because the format is standardized, Power Query can translate the column selection and row filtering you do in the editor into instructions the server itself understands, so the server does the work instead of your laptop. That translation is called query folding, and whether it happens is the single biggest factor in whether a refresh takes seconds or hours.
Underneath the dialog, Power Query is calling a function named OData.Feed. Microsoft's connector page names it directly, alongside a companion helper called ODataOmitValues.Nulls that matters a great deal more than its obscurity suggests.
Connecting a data source is the easy half. Knowing which numbers deserve a dashboard at all is the hard half. Get a free audit and find out which of your analytics and search metrics actually move the business.
Free audit
Where do you stand in AI search?
A manual SEO + AI search audit of your site: how ChatGPT, Gemini, Perplexity and Claude read and cite it today, and what to fix first. Report within 48 hours, no credit card.
Get my free auditOr have it all done: AI SEO services, $999 once
How Does an OData Feed Actually Work?
Three things decide whether a feed behaves: the address you point at, the query options that travel with it, and what the connector dialog lets you set.
The Service Root Is the Address You Need
Microsoft is specific about what to type. The URL "should be the root of the OData service you want to connect to," and the worked example on the page is a public test service: https://services.odata.org/V4/northwind/northwind.svc/.
That instruction sounds pedantic and is not. A root URL lets Power Query read the whole service description and show you every table in the Navigator. A deeper URL — one already pointing at a single table, or already carrying a filter — narrows what Power BI can see, and as the limitations below explain, it also breaks a refresh check in the Power BI service in a way that produces a misleading error.
The Query Options Are How You Filter at the Source
OData defines query options that travel in the URL, the familiar $select, $filter and $expand among them. These are what make a feed cheap to read. $select cuts columns, $filter cuts rows, and $expand pulls in related records in the same request.
You do not normally write them by hand. Microsoft's performance guidance for Business Central, whose connector "uses the OData.Feed library behind the scenes," spells out the translation Power Query performs: Table.RemoveColumns or Table.SelectColumns becomes $select, Table.SelectRows becomes $filter, and Table.ExpandTableColumn becomes $expand.
The same page shows the result as it reaches the service — a request carrying $filter=postingDate gt 2022-08-17 rather than a demand for the entire sales invoice table. That is the whole game: a filter step in the editor becoming a smaller question asked of the server.
What the Connector Dialog Actually Offers
Microsoft's capability list for the connector is short, and worth reading literally rather than skimming:
- Basic — a single URL box, which is what most connections need.
- Advanced — which exposes URL parts and open type columns, for services whose schema is not fully fixed.
- Select related tables — a button that pulls in any table having a direct relationship to one you already picked.
How Do You Connect Power BI to an OData Feed?
The documented path in Power BI Desktop runs to five steps:
- In the get data experience, choose OData or OData Feed. Microsoft's Azure DevOps guidance gives the full menu path as Other > OData Feed > Connect.
- Select Basic and enter the service root URL, then select OK.
- On a first connection, choose the authentication type, enter credentials if they are needed, and choose the level the authentication setting applies to. Then select Connect.
- In the Navigator dialog, select the table or tables you want. Use Select Related Tables if you also need the entities hanging off them.
- Choose Transform Data to land in Power Query, or Load to go straight to the model.
On authentication, the connector documents five kinds: Anonymous, Windows, Basic, Web API and Organizational Account. Two of them carry a condition worth noticing before you promise anyone a scheduled refresh, because Microsoft marks both Windows and Basic as "requires Gateway."
One difference if you are working in Power Query Online rather than Desktop: Microsoft's page offers only Transform Data from the Navigator there, with no Load option listed.
Choose Transform Data anyway, on either surface. Anything you can push into the query before loading is work the server does instead of your machine, and the editor is where folding either happens or quietly stops.
Most reporting problems are not connector problems. Get a free audit and see where your analytics and search data are pointing you in the wrong direction.
When Is an OData Feed the Right Choice?
A feed earns its place when the data lives in a system that publishes one and you want current numbers rather than a snapshot somebody exported last Thursday.
Microsoft's own products supply the obvious cases. Azure DevOps exposes Analytics as an OData feed at https://analytics.dev.azure.com/{OrganizationName}/_odata/{version}/. Business Central publishes APIs and web service endpoints the same way. SharePoint lists and Project Online are reachable over OData as well, which is why the connector page's limitations section keeps naming them.
It is the wrong choice when a purpose-built connector exists for the same system, because a dedicated connector usually knows the service's quirks, its pagination and its authentication better than a generic one can.
And it is the wrong choice for some APIs that look like ideal candidates. Microsoft states plainly that "Connecting to Microsoft Graph REST APIs from Power Query isn't recommended or supported," despite Graph being an OData-shaped API that many people reasonably assume would work.
What Are the Limitations Worth Knowing?
The OData connector has a reputation for being temperamental. Most of that reputation is documented behavior rather than mystery.
The 2,100-Character URL Ceiling
Microsoft gives exactly one hard number on the connector page: "the maximum URL length for these connections is approximately 2,100 characters. Exceeding the character limit results in a 401 error." It applies to SharePoint list, SharePoint online list and Project Online connections.
A 401 is an authentication status code, so the error you see has nothing to do with the actual cause, and you can lose an afternoon re-entering credentials that were never wrong. Microsoft's remedy is the same as its advice about the address: start "with the root OData endpoint and then navigate and filter inside Power Query."
Joins Can Kill the Connection Outright
Microsoft is blunt here: "Due to the architecture of OData and other web connectors, joins can be slow. A slow join can sometimes cause the source to fail the connection, resulting in an error such as 'An existing connection was forcibly closed by the remote host.'"
The documented fix is to buffer the smaller side of the merge. Microsoft recommends "buffering the smaller of the queries being merged to optimize the performance," applying Table.Buffer before the join, and the Business Central performance page repeats the same workaround for the same reason.
Folding Stops in More Places Than You Expect
Two documented gaps matter more than the rest. Microsoft's Business Central guidance states that "Query folding is NOT applied when loading data preview," so the editor's responsiveness tells you very little about what a real refresh will do.
The sharper one is on Microsoft's Azure DevOps page, which warns that the approach it documents "always pulls in all the data into Power BI. Most of the filtering that's specified in the Power Query Editor is done client-side." Filtering in the editor and filtering at the source can look identical on screen and differ by an order of magnitude in cost.
The Null-Value Trap That Gets Accounts Throttled
This one is specific, documented and genuinely expensive. For the Azure DevOps Analytics feed, Microsoft writes that "Power Query attempts to resolve null values as errors by generating an additional query for every null value it encounters. This action can result in thousands of queries, which can quickly exceed your usage threshold, beyond which your user account gets throttled."
The documented remedy is to tell the service to omit nulls and to pin the OData version, replacing [Implementation="2.0"] in the M query with [Implementation="2.0",OmitValues = ODataOmitValues.Nulls,ODataVersion = 4]. That is where the odd little ODataOmitValues.Nulls helper on the connector page earns its place.
Scheduled Refresh Tests the Wrong Thing
If your query carries query options, Microsoft's instruction for setting up refresh in the Power BI service is to select Skip Test Connection, because the service "tests the credentials but ignores any query options that were specified in the M query." The test can therefore pass or fail for reasons that have nothing to do with whether your query works.
Third-Party Authentication Often Just Does Not Work
For services outside Microsoft's own estate, the page is unusually candid. OAuth and Entra ID authentication "doesn't work in most cases," and the stated advice is to contact the service owner rather than to keep trying configurations.
Clean data in the model is worth nothing if the metrics are wrong. Get a free audit and let us check what your analytics are really telling you.
The Row Limit Microsoft Never Publishes
Search for an OData feed row limit and you will find confident numbers. None of them come from Microsoft.
Across the connector reference, the Azure DevOps Analytics guidance and the Business Central performance page, Microsoft states no row limit, no record cap and no import size limit for an OData feed. The only hard number in any of them is the 2,100-character URL ceiling, and that governs the request, not the data behind it.
What Microsoft documents instead is failure behavior, which is a different thing. The Azure DevOps page warns that choosing Snapshot entities on large repositories "causes tens to hundreds of millions of work items to fail to load" — a description of what breaks, not a published threshold. The Business Central page describes a transport-connection timeout, says the server "will cancel the call" when a query cannot finish inside the configured limit for web service calls, and gives no number for that limit either.
So the practical ceiling on an OData feed is whatever the publishing service tolerates, which varies by service, by tenant and by how much you asked for. Treat any specific row figure you read elsewhere as invented until a vendor page shows it to you.
OData Feed vs the Alternatives
Against a dedicated connector for the same system, the dedicated connector usually wins. It tends to understand the service's authentication, pagination and schema conventions, and Microsoft's own Business Central connector is a good illustration — it uses OData.Feed underneath while hiding the parts you would otherwise get wrong.
Against the generic Web connector, OData Feed wins whenever the service really is OData. The Web connector hands you a document to parse; the OData connector hands you tables, relationships and a Navigator, and it can fold your filters into the request.
Against calling a REST API yourself, OData Feed wins on effort and loses on control. Worth keeping straight: that is a different job from using Power BI's own REST API, which exists for managing Power BI programmatically rather than for getting data into it.
Against a file export, the feed wins on freshness and loses on predictability. A CSV someone dropped in a folder is slow and stale, but it does not throttle your account at 2 a.m. because a column turned out to be full of nulls.
Pulling the right data in is step one; knowing which numbers deserve attention is step two. Get a free audit and start from the metrics that count.
Final Thoughts
An OData feed is not an exotic data source. It is a URL that returns a table and understands the words $select and $filter, and that second property is the only reason Power Query can be clever with it at all.
Almost every problem people hit traces back to the same two mistakes: pointing at something deeper than the service root, and filtering in the editor while believing they are filtering at the source. Fix those two and most of the connector's reputation evaporates.
If you are connecting a specific system rather than learning the concept, our walkthrough of connecting Microsoft Project to Power BI shows the same OData machinery applied end to end, including what to do when the obvious route is closed.
And when a vendor page gives you no number, that is information rather than an invitation. The absence of a published row limit is the honest answer about how much data an OData feed can carry, and it is a better guide than any figure somebody made up to fill the silence.
Want this done for you?
SEO Stuff gets businesses cited by ChatGPT, Gemini, Perplexity and Claude, and ranking in Google. Start with the free audit, a call, or the package.
See where you stand
A free, manual SEO + AI search audit of your site, with a report within 48 hours. No credit card, no sales call.
Get my free auditTalk it through
A free 15-minute call with the founder about whether AI search is a meaningful opportunity for your business.
Free strategy callHave it done
The Done-For-You Package: audit, 10 pages of content, 3 DR50+ placements, dashboard. $999 once, delivered in 21 business days.
See the package

