Incremental Refresh and Re-Publishing from Desktop

[js1401]

Hi there,

I've recently started using Power BI and am still getting familiar with things.

I've established an incremental refresh to take snapshots of a small data set (a sharepoint list) so I can show my user how that data has changed over time. I've used this method and it seems to be working great:

https://blog.crossjoin.co.uk/2020/04/13/keep-the-existing-data-in-your-power-bi-dataset-and-add-new-...

The only issue I seem to be having is that when I republish the report from PBI Desktop to add some new visuals, I seem to be clearing the data set that incremental refresh has been steadily building up.

What am I doing wrong here?

Many thanks in advance 🙂 

[TomMartens · Accepted solution]

Hey @js1401 ,

next to what @Greg_Deckler already suggested, you might consider to separate the content development and the data model development into 2 pbix files.

use one pbix that just contains the data model, and the refresh policy.

use the 2nd pbix and connect to the power bi dataset, that has been published with the 1st pbix

This also allows to create measures in the 2nd pbix, the "content pbix".

If you are in the lucky situation and have Power Premium you can enable the XMLA read/write endpoint this will allow using the external tool ALM Toolkit to update the structural changes without affecting existing partitions.

You also have to be aware that the dataset will grow. So it can happen that you can download the pbix, but you will not be able to open the pbix in Power BI Desktop to make your changes.

Regards,
Tom

[Greg_Deckler]

@js1401 - Right, you need to follow this process:

1. Download PBIX from Service

2. Make updates

3. Republish

You are overwriting the dataset each time you publish from desktop, so make sure you have the most recent copy of the data.

[js1401]

Hi @Greg_Deckler - many thanks for your response and makes sense but when I tried what you suggested, I couldn't download the .pbix - I believe this is one of the drawbacks of using incremental refresh as the dataset may be huge (although mine isn't). 

I will either go with @TomMartens or maybe do the snapshotting/data history at source.

All the best 🙂

James