Power BI's Refresh button now has three choices. Which one to press when a column changed

What does Refresh data only do in Power BI?

Microsoft’s page explains that selecting Refresh in Power BI Desktop and the service always starts with a schema sync and is followed by a data refresh. A schema sync means Power BI looks at the source and updates its own list of columns and types to match. A data refresh just loads the rows again. These two processes ran together for years, so you never had to think about it.

Now you can expand the Refresh button on the Home ribbon or in the Data pane and pick one of three options. Refresh schema and data does both jobs, the same as before. Sync schema only updates the model to the source’s structure, like a column type change or a new column, and it loads no new rows. Refresh data only loads fresh data while keeping the schema your model currently has.

You can also go one level down and refresh a single table. Microsoft states that selecting an individual table lets you choose whether to refresh its schema, its data, or both. This matters when one table is big and slow and only a small lookup table changed. You no longer have to wait for the whole model to reload.

I read these pages on 27 October 2026 and I have not checked the menu on every kind of model. Microsoft’s own example is a Direct Lake model, which is the kind that reads tables straight from a Fabric lakehouse, and the page does not say the menu is limited to that kind. If your build of Power BI Desktop does not show the arrow, you should check that it is the August 2026 version or newer.

Which refresh option should I press when my source changed?

The right choice depends on what changed in the source, so it helps to look at the situation. I will explain how I decide. I base this on what Microsoft says each choice does and what it says a schema change does to a model.

The first row of the table is the most common day. The sheet got this week’s sales, nobody touched the headers, and you want the new rows in the report. Refresh data only is the cleanest choice for this. It cannot surprise you with a structure change you did not plan.

The third row is the case Microsoft wrote the feature for. Someone adds a column to the source for their own use, and you do not want it turning up in your model, your field list, or an underlying data export that a viewer takes out of the report. You can use Refresh data only to get the new rows and leave the new column outside.

What happens to my report when a source column is renamed?

This next part catches people out, and Microsoft states it plainly on the same page. A schema refresh will remove a renamed or removed column from the model. This can break visuals and DAX expressions, which means your measures, calculated columns, and row-level security rules. It can also remove relationships that depend on that column or table.

I can give a simple example. Your model has a measure called Total Sales that adds up a column called Amount, and someone renames Amount to Net in the source sheet. A schema sync sees that Amount is gone and removes it from the model. Now Total Sales points at a column that does not exist, so every chart that uses it will show an error.

The service behaves differently, and you should know about it. Microsoft says a data refresh in the Power BI service will fail when a source column or table is renamed or removed, because the service does not do a schema refresh there. Their fix is to do the schema refresh in Power BI Desktop and publish the semantic model again. A semantic model is the data model behind the report, which includes the tables, relationships, and measures that the report reads from.

That same page still has an older table which says that the data refresh and the schema refresh are not available independently in Power BI Desktop. The newer section on refresh options says the exact opposite. I read that as a table that Microsoft has not updated yet. But it is one more reason to look at your own Refresh button before you trust any guide, and that includes mine.

Why does Refresh data only not fix a renamed column in Excel or CSV?

Most small business reports I see are import models built on an Excel file or a CSV through Power Query. Power Query is the tool inside Power BI that cleans the data before it loads, and it writes every step down with the column names in it. A step such as Changed Type lists each column by name. So when the source loses a column called Amount, that step stops with an error saying it cannot find the column.

No refresh choice fixes that issue, because the problem sits in the query and not in the model. Refresh data only still runs the query, and the query still asks for Amount. So the fix for a renamed column is in Power Query. You open Transform data, find the step that shows the error, and either rename the column back in a step of its own or change the step to the new name.

A CSV file has one more quirk that looks like a refresh problem. When Power Query opens a CSV, the first step usually writes down how many columns the file had. This means a column added later at the end of the file can be left out even by a full refresh. If a new CSV column never shows up, you should check the Source step before you blame the button.

How do I update a report safely after someone changed the source?

This is the order I would follow when a source sheet has changed and the report has to keep working. It assumes you have the report open in Power BI Desktop.

1. 01
   Find out exactly what changed

   Open the source and compare the headers with yesterday. A new column, a renamed one, a removed one and a changed type each need a different button.
2. 02
   If only rows changed, press Refresh data only

   Expand the Refresh button on the Home ribbon and pick Refresh data only. Then look at one total you already know.
3. 03
   If a column was renamed or removed, fix the query and the measures first

   Open Transform data, fix the step that errors, then search your measures and relationships for the old name. Only then refresh the schema.
4. 04
   If one table changed, refresh that table alone

   Pick the table in the Data pane and choose schema, data or both for it. The rest of the model stays as it was.
5. 05
   Publish again if the report lives in the service

   Microsoft says a structure change has to be applied in Power BI Desktop and the model published again. A scheduled refresh in the service will keep failing until you do.

Step two has a check at the end for a reason. A refresh that finishes without an error is not the same as a refresh that loaded the right rows. The only way to know is to look at a number you can check against the source. I made the same point about Excel in why Power Automate says it refreshed when nothing changed, and it holds just as well here.

Does pressing Refresh by hand use up my eight refreshes a day?

No. Microsoft limits each semantic model to eight scheduled refreshes a day on shared capacity, which is the normal Pro workspace. This count resets at 12:01 AM in the time zone set on the model. The same page says that manual, UI-triggered on-demand refreshes are not included in that limit.

The limit does cover refreshes started through the REST API. This means a flow in Power Automate that refreshes the model through the API uses up the same eight. The schedule can have up to 48 time slots on Premium, Premium Per User, or Fabric capacity. And a refresh you press in Power BI Desktop runs on your own PC against your own copy, so it has nothing to do with the service’s count at all.

So trying Refresh data only a few times while you sort out a changed source costs you nothing in the service. You should avoid the habit of adding scheduled times to fix a refresh that fails. A schedule cannot fix a renamed column, and it just fails eight times instead of once.