TR1901 - Base measure for category count cannot be calulated 1

[Rolldice]

Hello all,

I am building a CRM cube which merges 4 fact tables together (I know this isn't the best idea). But the client wants all the metrics in one cube and some of the measures are calulated across the differant CRM activity fact tables. So I have come up with a cube which was working very well. With each fact table being a new data source, that was until I added the last of the 4 data sources.

Until this point I had a category count giving me a nice distinct count on customer ID's, in the customer dimension.

However once I added the last data scorce and ran the build, which takes about 2 hours the measure with the category count was zero in the cube. I checked the scope of the measure and it still looked as good as before I added the last data source. Having run the cube again I recieve the TR1901 error message: The base measure for category count cannot be calulated affter rollup, allocated or external.

Some of the other fact tables have the cust_id field, but they are grouped differently could this be coursing the issue?

I looked at a few treads and TR1901 looks to be a memory issue, however I have been building this cube previously to the new data source.

Any ideas?

----------------------------------------
Another throw closer to a win ... ???
----------------------------------------

 

[yksvaan]

Look at the category count measure's Rollup tab. What do you have there as a Activity measure? Have you maybe excluded some measures from cube? What does Tools -> Check model say ?

Cheers


[blue]Backup system is as good as the latest recovery[/blue]

 

[Rolldice]

yksvann - Thanks for your reply,

In the category count measures rollup Tab, the activity measure is set to All measures.

Should I set this to just the measure itself? What is the "all Measures" activity measure meaning for this category count... I think we are on the right track.

Should I also de-allocate the distinct count measure from all the dimension properties?

The check model is fine - No errors.


----------------------------------------
Another throw closer to a win ... ???
----------------------------------------

 

[yksvaan]

Quick answer: Don't do any of those. I'll get back.

Cheers


[blue]Backup system is as good as the latest recovery[/blue]

 

[yksvaan]

Have you really explicitly allocated the customer count measure to some dimensions(?) because you mentioned de-allocate, I didn't follow on that? When Activity measure is All measures, you could try change it to a measure, that is related to Customer Id. Activity measure means that if the measures value is zero or missing, that customer won't be counted as 1.

Or you might have this situation, check this first anyway:
---
The base measure was selected to be allocated in the dimension but this could not be achieved as the lowest level was out of scope.
This level was out of scope as it's data column was only in one of the two datasources whereas it's parental level datasource was in each of the data sources.
Issue addressed by ensuring that the lowest level data column exists in the same data source as it's parental level data column.
---
What that says is that make sure when you are referring to a dimension in a query, that you have exactly the same columns concerning that hierarcy in all the other queries referring to this dimension.

Cheers


[blue]Backup system is as good as the latest recovery[/blue]

 

[Rolldice]

yksvaan,

Cheers for that reply - I beleive that I have the second situation in my case. It looks like I have created an unbalanced category tree for the customer dimension due to the different data scores coming into the model with cust_id.

In the properties tab of the datascores I have adjusted the timing so that only the relevant datasource create the catagories for the customer dimension. This has evened the customer dimension up and the cube is now working.

Regarding the de-allocation experimentation - I was changing the state of the Allocation type in the dimension porperties, while checking the scope of each measure.. If that makes sense?? There's the option of: Do not allocate / Constant / by measure. I wasn't quite up on what that would achieve but I gave it a go in vain.

Thanks again - Here's a star for U!




----------------------------------------
Another throw closer to a win ... ???
----------------------------------------